Math 222 Second Semester Calculus

MATH 222 SECOND SEMESTER CALCULUS

Spring 2011

1 2

Math 222 – 2nd Semester Calculus Lecture notes version 1.7(Spring 2011) This is a self contained set of lecture notes for Math 222. The notes were written by Sigurd Angenent, starting from an extensive collection of notes and problems compiled by Joel Robbin. Some problems were contributed by A.Miller. The LATEX ﬁles, as well as the Xfig and Octave ﬁles which were used to produce these notes are available at the following web site www.math.wisc.edu/~angenent/Free-Lecture-Notes They are meant to be freely available for non-commercial use, in the sense that “free software” is free. More precisely:

Copyright (c) 2006 Sigurd B. Angenent. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled ”GNU Free Documentation License”.

Contents

Chapter 1: Methods of Integration 3 1. The indeﬁnite integral 3 2. You can always check the answer 4 3. About “+C” 4 4. Standard Integrals 5 5. Method of substitution 5 6. The double angle trick 7 7. Integration by Parts 7 8. Reduction Formulas 9 9. Partial Fraction Expansion 12 10. PROBLEMS 16

Chapter 2: Taylor’s Formulaand Infinite Series 27 11. Taylor Polynomials 27 12. Examples 28 13. Some special Taylor polynomials 32 14. The Remainder Term 32 15. Lagrange’s Formula for the Remainder Term 34 16. The limit as x 0, keeping n fixed 36 → 17. The limit n , keeping x fixed 43 →∞ 18. Convergence of Taylor Series 46 19. Leibniz’ formulas for ln 2 and π/4 48 20. Proof of Lagrange’s formula 49 21. Proof of Theorem 16.8 50 22. PROBLEMS 51

Chapter 3: Complex Numbers and the Complex Exponential 56 23. Complex numbers 56 24. Argument and Absolute Value 57 25. Geometry of Arithmetic 58 26. Applications in Trigonometry 60 27. Calculus of complex valued functions 61 3

28. The Complex Exponential Function 61 29. Complex solutions of polynomial equations 63 30. Other handy things you can do with complex numbers 65 31. PROBLEMS 67

Chapter 4: Differential Equations 72 32. What is a DiffEq? 72 33. First Order Separable Equations 72 34. First Order Linear Equations 73 35. Dynamical Systems and Determinism 75 36. Higher order equations 77 37. Constant Coefficient Linear Homogeneous Equations 78 38. Inhomogeneous Linear Equations 81 39. Variation of Constants 82 40. Applications of Second Order Linear Equations 85 41. PROBLEMS 89

Chapter 5: Vectors 97 42. Introduction to vectors 97 43. Parametric equations for lines and planes 102 44. Vector Bases 104 45. Dot Product 105 46. Cross Product 112 47. A few applications of the cross product 115 48. Notation 118 49. PROBLEMS 118

Chapter 6: Vector Functions and Parametrized Curves 124 50. Parametric Curves 124 51. Examples of parametrized curves 125 52. The derivative of a vector function 127 53. Higher derivatives and product rules 128 54. Interpretation of x′(t) as the velocity vector 129 55. Acceleration and Force 131 56. Tangents and the unit tangent vector 133 57. Sketching a parametric curve 135 58. Length of a curve 137 59. The arclength function 139 60. Graphs in Cartesian and in Polar Coordinates 140 61. PROBLEMS 141

GNU Free Documentation License 148 1. APPLICABILITY AND DEFINITIONS 148 2. VERBATIM COPYING 149 3. COPYING IN QUANTITY 149 4. MODIFICATIONS 149 5. COMBINING DOCUMENTS 150 6. COLLECTIONS OF DOCUMENTS 150 7. AGGREGATION WITH INDEPENDENT WORKS 150 8. TRANSLATION 150 9. TERMINATION 150 10. FUTURE REVISIONS OF THIS LICENSE 151 4

11. RELICENSING 151 5

Chapter 1: Methods of Integration

1. The indeﬁnite integral

We recall some facts about integration from ﬁrst semester calculus.

1.1. Deﬁnition. A function y = F (x) is called an antiderivative of another function y = f(x) if F ′(x)= f(x) for all x.

2 1.2. Example. F1(x)= x is an antiderivative of f(x)=2x. 2 F2(x)= x + 2004 is also an antiderivative of f(x)=2x. 1 G(t)= 2 sin(2t + 1) is an antiderivative of g(t) = cos(2t + 1). The Fundamental Theorem of Calculus states that if a function y = f(x) is continuous on an interval a x b, then there always exists an antiderivative F (x) of f, and one has ≤ ≤ b (1) f(x) dx = F (b) F (a). Za − The best way of computing an integral is often to ﬁnd an antiderivative F of the given function f, and then to use the Fundamental Theorem (1). How you go about ﬁnding an antiderivative F for some given function f is the subject of this chapter. The following notation is commonly used for antiderivates:

(2) F (x)= f(x)dx. Z The integral which appears here does not have the integration bounds a and b. It is called an indefinite integral, as opposed to the integral in (1) which is called a definite integral. It’s important to distinguish between the two kinds of integrals. Here is a list of differences: Indefinite integral Definite integral

b f(x)dx is a function of x. a f(x)dx is a number. R b R By definition f(x)dx is any func- a f(x)dx was defined in terms of tion of x whoseR derivative is f(x). RiemannR sums and can be interpreted as “area under the graph of y = f(x)”, at least when f(x) > 0. x is not a dummy variable, for exam- x is a dummy variable, for example, ple, 2xdx = x2 + C and 2tdt = 1 1 0 2xdx = 1, and 0 2tdt = 1, so t2 +CRare functions of diffferentR vari- R 1 1 R 0 2xdx = 0 2tdt. ables, so they are not equal. R R 6

2. You can always check the answer

Suppose you want to ﬁnd an antiderivative of a given function f(x) and after a long and messy computation which you don’t really trust you get an “answer”, F (x). You can then throw away the dubious computation and diﬀerentiate the F (x) you had found. If F ′(x) turns out to be equal to f(x), then your F (x) is indeed an antiderivative and your computation isn’t important anymore.

2.1. Example. Suppose we want to ﬁnd ln x dx. My cousin Bruce says it might be F (x)= x ln x x. Let’s see if he’s right:R − d 1 (x ln x x)= x +1 ln x 1 = ln x. dx − x − Who knows how Bruce thought of this1, but he’s right! We now know that ln xdx = x ln x x + C. − R 3. About “+C”

Let f(x) be a function defined on some interval a x b. If F (x) is an ≤ ≤ antiderivative of f(x) on this interval, then for any constant C the function F˜(x)= F (x)+ C will also be an antiderivative of f(x). So one given function f(x) has many different antiderivatives, obtained by adding different constants to one given antiderivative.

3.1. Theorem. If F1(x) and F2(x) are antiderivatives of the same function f(x) on some interval a x b, then there is a constant C such that F (x) = ≤ ≤ 1 F2(x)+ C. Proof. ′ ′ Consider the diﬀerence G(x)= F1(x) F2(x). Then G (x)= F1(x) ′ − − F2(x)= f(x) f(x) = 0, so that G(x) must be constant. Hence F1(x) F2(x)= C for some constant.− −

It follows that there is some ambiguity in the notation f(x) dx. Two functions F1(x) and F2(x) can both equal f(x) dx without equalingR each other. When this happens, they (F1 and F2) diﬀerR by a constant. This can sometimes lead to confusing situations, e.g. you can check that

2 sin x cos x dx = sin2 x Z 2 sin x cos x dx = cos2 x Z − are both correct. (Just differentiate the two functions sin2 x and cos2 x!) These − two answers look different until you realize that because of the trig identity sin2 x+ cos2 x = 1 they really only differ by a constant: sin2 x = cos2 x + 1. − To avoid this kind of confusion we will from now on never forget to include the “arbitrary constant +C” in our answer when we compute an antiderivative.

1He integrated by parts. 7

4. Standard Integrals

Here is a list of the standard derivatives and hence the standard integrals everyone should know.

f(x) dx = F (x)+ C Z xn+1 xn dx = + C for all n = 1 Z n +1 − 1 dx = ln x + C Z x | | sin x dx = cos x + C Z − cos x dx = sin x + C Z tan x dx = ln cos x + C Z − 1 dx = arctan x + C Z 1+ x2 1 π dx = arcsin x + C (= arccos x + C) Z √1 x2 2 − − dx 1 1 + sin x π π = ln + C for

5. Method of substitution

The chain rule says that

dF (G(x)) ′ ′ = F (G(x)) G (x), dx so that ′ ′ F (G(x)) G (x) dx = F (G(x)) + C. Z

5.1. Example. Consider the function f(x)= 2x sin(x2 +3). It does not appear in the list of standard integrals we know by heart. But we do notice2 that 2x = d (x2 + 3). So let’s call G(x)= x2 + 3, and F (u)= cos u, then dx − F (G(x)) = cos(x2 + 3) − and dF (G(x)) = sin(x2 + 3) 2x = f(x), dx ′ F ′(G(x)) G (x) | {z } |{z} 2 You will start noticing things like this after doing several examples. 8 so that 2x sin(x2 +3)dx = cos(x2 +3)+ C. Z − The most transparent way of computing an integral by substitution is by in- troducing new variables. Thus to do the integral

′ f(G(x))G (x) dx Z where f(u) = F ′(u), we introduce the substitution u = G(x), and agree to write du = dG(x)= G′(x) dx. Then we get

′ f(G(x))G (x) dx = f(u) du = F (u)+ C. Z Z At the end of the integration we must remember that u really stands for G(x), so that ′ f(G(x))G (x) dx = F (u)+ C = F (G(x)) + C. Z For deﬁnite integrals this implies b ′ f(G(x))G (x) dx = F (G(b)) F (G(a)). Za − which you can also write as

b G(b) ′ (3) f(G(x))G (x) dx = f(u) du. Za ZG(a)

5.2. Example. [Substitution in a definite integral. ] As an example we compute 1 x 2 dx, Z0 1+ x using the substitution u = G(x) =1+ x2. Since du = 2x dx, the associated indefinite integral is 1 1 x dx = 1 du. Z 1+ x2 2 Z u 1 du 1 2 u |{z} | {z } To find the definite integral you must compute the new integration bounds G(0) and G(1) (see equation (3).) If x runs between x = 0 and x = 1, then u = G(x)=1+x2 runs between u =1+02 = 1 and u =1+12 = 2, so the definite integral we must compute is 1 2 x 1 1 2 dx = 2 du, Z0 1+ x Z1 u which is in our list of memorable integrals. So we find 1 2 x 1 2 dx = 1 du = 1 ln u = 1 ln 2. Z 1+ x2 2 Z u 2 1 2 0 1 9

6. The double angle trick

If an integral contains sin2 x or cos2 x, then you can remove the squares by using the double angle formulas from trigonometry. Recall that cos2 α sin2 α = cos2α and cos2 α + sin2 α =1, − Adding these two equations gives 1 cos2 α = (cos2α + 1) 2 while substracting them gives 1 sin2 α = (1 cos2α) . 2 −

6.1. Example. The following integral shows up in many contexts, so it is worth knowing: 1 cos2 x dx = (1+cos2x)dx Z 2 Z 1 1 = x + sin 2x + C 2 2 x 1 = + sin 2x + C. 2 4 Since sin 2x = 2 sin x cos x this result can also be written as x 1 cos2 x dx = + sin x cos x + C. Z 2 2

If you don’t want to memorize the double angle formulas, then you can use “Complex Exponentials” to do these and many similar integrals. However, you will have to wait until we are in 28 where this is explained. §

7. Integration by Parts

The product rule states d dF (x) dG(x) (F (x)G(x)) = G(x)+ F (x) dx dx dx and therefore, after rearranging terms, dG(x) d dF (x) F (x) = (F (x)G(x)) G(x). dx dx − dx This implies the formula for integration by parts dG(x) dF (x) F (x) dx = F (x)G(x) G(x) dx. Z dx − Z dx 10

7.1. Example – Integrating by parts once.

x ex dx = x ex ex 1 dx = xex ex + C. Z − Z − F (x) G′(x) F (x) G(x) G(x) F ′(x) |{z} |{z} |{z} |{z} |{z} |{z} Observe that in this example ex was easy to integrate, while the factor x becomes an easier function when you differentiate it. This is the usual state of affairs when integration by parts works: differentiating one of the factors (F (x)) should simplify the integral, while integrating the other (G′(x)) should not complicate things (too much). Another example: sin x = d ( cos x) so dx − x sin x dx = x( cos x) ( cos x) 1 dx == x cos x + sin x + C. Z − − Z − −

7.2. Example – Repeated Integration by Parts. Sometimes one integra- 2x d 1 2x tion by parts is not enough: since e = dx ( 2 e ) one has e2x e2x x2 e2x dx = x2 2x dx Z 2 − Z 2 F (x) G′(x) |{z} |{z} e2x e2x e2x = x2 2x 2 dx 2 − 4 − Z 4 e2x e2x e2x = x2 2x 2+ C 2 − 4 − 8 1 1 1 = x2e2x xe2x + e2x C 2 − 2 4 − (Be careful with all the minus signs that appear when you integrate by parts.) The same procedure will work whenever you have to integrate

P (x)eax dx Z where P (x) is a polynomial, and a is a constant. Each time you integrate by parts, you get this ax ax e e ′ P (x)eax dx = P (x) P (x) dx Z a − Z a 1 1 ′ = P (x)eax P (x)eax dx. a − a Z You have replaced the integral P (x)eax dx with the integral P ′(x)eax dx. This is the same kind of integral, butR it is a little easier since the degreeR of the derivative P ′(x) is less than the degree of P (x).

7.3. Example – My cousin Bruce’s computation. Sometimes the factor G′(x) is “invisible”. Here is how you can get the antiderivative of ln x by integrating 11 by parts:

ln x dx = ln x 1 dx Z Z F (x) G′(x) |{z} |{z}1 = ln x x x dx − Z x = x ln x 1 dx − Z = x ln x x + C. − You can do P (x) ln x dx in the same way if P (x) is a polynomial. R 8. Reduction Formulas

Consider the integral n ax In = x e dx. Z Integration by parts gives you

n 1 ax n−1 1 ax In = x e nx e dx a − Z a 1 n − = xneax xn 1eax dx. a − a Z We haven’t computed the integral, and in fact the integral that we still have to do is of the same kind as the one we started with (integral of xn−1eax instead of xneax). What we have derived is the following reduction formula

1 n ax n In = x e In− (R) a − a 1 which holds for all n. For n = 0 the reduction formula says

1 ax ax 1 ax I0 = e , i.e. e dx = e + C. a Z a

When n = 0 the reduction formula tells us that we have to compute In− if we 1 want to ﬁnd In. The point of a reduction formula is that the same formula also applies to In−1, and In−2, etc., so that after repeated application of the formula we end up with I0, i.e., an integral we know.

8.1. Example. To compute x3eax dx we use the reduction formula three times: R 1 3 I = x3eax I 3 a − a 2 1 3 1 2 = x3eax x2eax I a − a a − a 1 1 3 1 2 1 1 = x3eax x2eax xeax I a − a a − a a − a 0 12

1 ax Insert the known integral I0 = a e + C and simplify the other terms and you get 1 3 6 6 x3eax dx = x3eax x2eax + xeax eax + C. Z a − a2 a3 − a4

8.2. Reduction formula requiring two partial integrations. Consider

n Sn = x sin x dx. Z Then for n 2 one has ≥ n n−1 Sn = x cos x + n x cos x dx − Z − − = xn cos x + nxn 1 sin x n(n 1) xn 2 sin x dx. − − − Z Thus we ﬁnd the reduction formula n n−1 Sn = x cos x + nx sin x n(n 1)Sn− . − − − 2 Each time you use this reduction, the exponent n drops by 2, so in the end you get either S1 or S0, depending on whether you started with an odd or even n.

8.3. A reduction formula where you have to solve for In. We try to compute n In = (sin x) dx Z by a reduction formula. Integrating by parts twice we get

n−1 In = (sin x) sin x dx Z − − = (sin x)n 1 cos x ( cos x)(n 1)(sin x)n 2 cos x dx − − Z − − − − = (sin x)n 1 cos x + (n 1) (sin x)n 2 cos2 x dx. − − Z We now use cos2 x =1 sin2 x, which gives − n−1 n−2 n In = (sin x) cos x + (n 1) sin x sin x dx − − Z − n−1 = (sin x) cos x + (n 1)In− (n 1)In. − − 2 − − You can think of this as an equation for In, which, when you solve it tells you n−1 nIn = (sin x) cos x + (n 1)In− − − 2 and thus implies 1 n−1 n 1 In = sin x cos x + − In− . (S) −n n 2 Since we know the integrals

0 I0 = (sin x) dx = dx = x + C and I1 = sin x dx = cos x + C Z Z Z − the reduction formula (S) allows us to calculate In for any n 0. ≥ 13

8.4. A reduction formula which will be handy later. In the next section you will see how the integral of any “rational function” can be transformed into integrals of easier functions, the hardest of which turns out to be

dx In = . Z (1 + x2)n

When n = 1 this is a standard integral, namely

dx I1 = = arctan x + C. Z 1+ x2 When n > 1 integration by parts gives you a reduction formula. Here’s the computation:

2 −n In = (1 + x ) dx Z x −n−1 = x ( n) 1+ x2 2x dx (1 + x2)n − Z − x x2 = +2n dx (1 + x2)n Z (1 + x2)n+1

Apply x2 (1 + x2) 1 1 1 = − = (1 + x2)n+1 (1 + x2)n+1 (1 + x2)n − (1 + x2)n+1 to get

x2 1 1 dx = dx = In In+1. Z (1 + x2)n+1 Z (1 + x2)n − (1 + x2)n+1 −

Our integration by parts therefore told us that x In = +2n In In , (1 + x2)n − +1 which you can solve for In+1. You ﬁnd the reduction formula 1 x 2n 1 I = + − I . n+1 2n (1 + x2)n 2n n

As an example of how you can use it, we start with I1 = arctan x + C, and conclude that dx = I2 = I1+1 Z (1 + x2)2 1 x 2 1 1 = + − I 2 1 (1 + x2)1 2 1 1 x = 1 + 1 arctan x + C. 2 1+ x2 2 14

Apply the reduction formula again, now with n = 2, and you get dx = I3 = I2+1 Z (1 + x2)3 1 x 2 2 1 = + − I 2 2 (1 + x2)2 2 2 2 x x = 1 + 3 1 + 1 arctan x 4 (1 + x2)2 4 2 1+ x2 2 x x = 1 + 3 + 3 arctan x + C. 4 (1 + x2)2 8 1+ x2 8

9. Partial Fraction Expansion

A rational function is one which is a ratio of polynomials, n n−1 P (x) pnx + pn−1x + + p1x + p0 f(x)= = d d−1 . Q(x) qdx + qd− x + + q x + q 1 1 0 Such rational functions can always be integrated, and the trick which allows you to do this is called a partial fraction expansion. The whole procedure consists of several steps which are explained in this section. The procedure itself has nothing to do with integration: it’s just a way of rewriting rational functions. It is in fact useful in other situations, such as ﬁnding Taylor series (see Part 178 of these notes) and computing “inverse Laplace transforms” (see Math 319.)

9.1. Reduce to a proper rational function. A proper rational function is a rational function P (x)/Q(x) where the degree of P (x) is strictly less than the degree of Q(x). the method of partial fractions only applies to proper rational functions. Fortunately there’s an additional trick for dealing with rational functions that are not proper. If P/Q isn’t proper, i.e. if degree(P ) degree(Q), then you divide P by Q, with result ≥ P (x) R(x) = S(x)+ Q(x) Q(x) where S(x) is the quotient, and R(x) is the remainder after division. In practice you would do a long division to ﬁnd S(x) and R(x).

9.2. Example. Consider the rational function x3 2x +2 f(x)= − . x2 1 − Here the numerator has degree 3 which is more than the degree of the denominator (which is 2). To apply the method of partial fractions we must ﬁrst do a division with remainder. One has x +1 = S(x) x2 1 x3 2x +2 − x3 − x −x +2 = R(x) − so that x3 2x +2 x +2 f(x)= − = x + − x2 1 x2 1 − − 15

When we integrate we get x3 2x +2 x +2 − dx = x + − dx Z x2 1 Z x2 1 − − x2 x +2 = + − dx. 2 Z x2 1 − −x+2 The rational function which still have to integrate, namely x2−1 , is proper, i.e. its numerator has lower degree than its denominator.

9.3. Partial Fraction Expansion: The Easy Case. To compute the partial fraction expansion of a proper rational function P (x)/Q(x) you must factor the denominator Q(x). Factoring the denominator is a problem as difficult as finding all of its roots; in Math 222 we shall only do problems where the denominator is already factored into linear and quadratic factors, or where this factorization is easy to find. In the easiest partial fractions problems, all the roots of Q(x) are real numbers and distinct, so the denominator is factored into distinct linear factors, say P (x) P (x) = . Q(x) (x a )(x a ) (x an) − 1 − 2 − To integrate this function we find constants A1, A2,...,An so that P (x) A A A = 1 + 2 + + n . (#) Q(x) x a x a x an − 1 − 2 − Then the integral is P (x) dx = A1 ln x a1 + A2 ln x a2 + + An ln x an + C. Z Q(x) | − | | − | | − |

One way to find the coefficients Ai in (#) is called the method of equating coefficients. In this method we multiply both sides of (#) with Q(x) = (x − a ) (x an). The result is a polynomial of degree n on both sides. Equating 1 − the coefficients of these polynomial gives a system of n linear equations for A1,..., An. You get the Ai by solving that system of equations. 3 Another much faster way to find the coefficients Ai is the Heaviside trick . 4 Multiply equation (#) by x ai and then plug in x = ai. On the right you are − left with Ai so

P (x)(x ai) P (ai) Ai = − = . Q(x) (a a ) (a a − )(a a ) (a a ) x=ai i 1 i i 1 i i+1 i n − − − −

3 Named after Oliver Heaviside, a physicist and electrical engineer in the late 19th and early 20ieth century. 4 More properly, you should take the limit x a . The problem here is that equation (#) → i has x a in the denominator, so that it does not hold for x = a . Therefore you cannot set x − i i equal to a in any equation derived from (#), but you can take the limit x a , which in practice i → i is just as good. 16

x +2 9.4. Previous Example continued. To integrate − we factor the de- x2 1 nominator, − x2 1 = (x 1)(x + 1). − − x +2 The partial fraction expansion of − then is x2 1 − x +2 x +2 A B − = − = + . ( ) x2 1 (x 1)(x + 1) x 1 x +1 † − − − Multiply with (x 1)(x +1) to get − x +2= A(x +1)+ B(x 1)=(A + B)x + (A B). − − − The functions of x on the left and right are equal only if the coefficient of x and the constant term are equal. In other words we must have A + B = 1 and A B =2. − − These are two linear equations for two unknowns A and B, which we now proceed to 1 solve. Adding both equations gives 2A = 1, so that A = 2 ; from the first equation one then finds B = 1 A = 3 . So − − − 2 x +2 1/2 3/2 − = . x2 1 x 1 − x +1 − − Instead, we could also use the Heaviside trick: multiply ( ) with x 1 to get † − x +2 x 1 − = A + B − x +1 x +1 Take the limit x 1 and you find → 1+2 1 − = A, i.e. A = . 1+1 2 Similarly, after multiplying ( ) with x +1 one gets † x +2 x +1 − = A + B, x 1 x 1 − − and letting x 1 you find →− ( 1)+2 3 B = − − = , ( 1) 1 −2 − − as before. Either way, the integral is now easily found, namely, x3 2x +1 x2 x +2 − dx = + x + − dx Z x2 1 2 Z x2 1 − − x2 1/2 3/2 = + x + dx 2 Z x 1 − x +1 − x2 1 3 = + x + ln x 1 ln x +1 + C. 2 2 | − |− 2 | | 17

9.5. Partial Fraction Expansion: The General Case. Buckle up. When the denominator Q(x) contains repeated factors or quadratic factors (or both) the partial fraction decomposition is more complicated. In the most general case the denominator Q(x) can be factored in the form

k1 kn 2 ℓ1 2 ℓm (4) Q(x) = (x a ) (x an) (x + b x + c ) (x + bmx + cm) − 1 − 1 1 Here we assume that the factors x a1,..., x an are all diﬀerent, and we also 2 − 2 − assume that the factors x + b1x + c1,..., x + bmx + cm are all diﬀerent. It is a theorem from advanced algebra that you can always write the rational function P (x)/Q(x) as a sum of terms like this P (x) A Bx + C (5) = + k + + 2 ℓ + Q(x) (x ai) (x + bjx + cj ) − How did this sum come about? For each linear factor (x a)k in the denominator (4) you get terms − A A A 1 + 2 + + k x a (x a)2 (x a)k − − − in the decomposition. There are as many terms as the exponent of the linear factor that generated them. For each quadratic factor (x2 + bx + c)ℓ you get terms B x + C B x + C B x + C 1 1 + 2 2 + + m m . x2 + bx + c (x2 + bx + c)2 (x2 + bx + c)ℓ Again, there are as many terms as the exponent ℓ with which the quadratic factor appears in the denominator (4).

In general, you ﬁnd the constants A..., B... and C... by the method of equating coeﬃcients.

9.6. Example. To do the integral x2 +3 dx Z x2(x + 1)(x2 + 1)2 apply the method of equating coefficients to the form x2 +3 A A A B x + C B x + C = 1 + 2 + 3 + 1 1 + 2 2 . (EX) x2(x + 1)(x2 + 1)2 x x2 x +1 x2 +1 (x2 + 1)2 Solving this last problem will require solving a system of seven linear equations in the seven unknowns A1, A2, A3,B1, C1,B2, C2. A computer program like Maple can do this easily, but it is a lot of work to do it by hand. In general, the method of equating coefficients requires solving n linear equations in n unknowns where n is the degree of the denominator Q(x). See Problem 104 for a worked example where the coefficients are found.

Unfortunately, in the presence of quadratic factors or repeated lin- !! ear factors the Heaviside trick does not give the whole answer; you !! must use the method of equating coeﬃcients. 18

Once you have found the partial fraction decomposition (EX) you still have to integrate the terms which appeared. The ﬁrst three terms are of the form A(x a)−p dx and they are easy to integrate: − R A dx = A ln x a + C Z x a | − | − and A dx A = + C Z (x a)p (1 p)(x a)p−1 − − − if p> 1. The next, fourth term in (EX) can be written as

B1x + C1 x dx dx = B1 dx + C1 Z x2 +1 Z x2 +1 Z x2 +1 B = 1 ln(x2 +1)+ C arctan x + C 2 1 integration const. While these integrals are already not very simple, the integrals Bx + C dx with p> 1 Z (x2 + bx + c)p which can appear are particularly unpleasant. If you really must compute one of these, then complete the square in the denominator so that the integral takes the form Ax + B dx. Z ((x + b)2 + a2)p After the change of variables u = x + b and factoring out constants you have to do the integrals du u du and . Z (u2 + a2)p Z (u2 + a2)p Use the reduction formula we found in example 8.4 to compute this integral. An alternative approach is to use complex numbers (which are on the menu for this semester.) If you allow complex numbers then the quadratic factors x2 + bx + c can be factored, and your partial fraction expansion only contains terms of the form A/(x a)p, although A and a can now be complex numbers. The integrals are then easy,− but the answer has complex numbers in it, and rewriting the answer in terms of real numbers again can be quite involved.

10. PROBLEMS

DEFINITE VERSUS INDEFINITE INTEGRALS

1. Compute the following three integrals:

2 2 2 A = x− dx, B = t− dt, C = x− dt. 2. One of the following three integrals is not the same as the other two: 4 4 4 2 2 2 A = x− dx, B = t− dt, C = x− dt. 1 1 1 Which one? BASIC INTEGRALS 19

The following integrals are straightforward provided you know the list of standard antiderivatives. They can be done without using substitution or any other tricks, and you learned them in ﬁrst semester calculus.

π/3 5 4 3. 6x 2x− 7x − − 22. sin tdt x x π/4 +3/x 5 + 4e + 7 dx − π/2 a x 23. (cos θ + 2 sin θ) dθ 4. (x/a + a/x + x + a + ax) dx 0 π/2 3 7 5. √x √x4 + 6ex + 1 dx 24. (cos θ + sin 2θ) dθ − √3 2 − 0 x π tan x 6. 2x + 1 x dx 25. dx 2 2π/3 cos x 0 π/2 cot x 7. (5y4 6y2 + 14) dy 26. dx 3 − π/3 sin x − 3 1 1 √3 6 8. dt 27. dx t2 − t4 1+ x2 1 1 2 t6 t2 0.5 dx 9. − dt 28. 4 2 1 t 0 √1 x − 2 x2 + 1 8 10. dx 29. (1/x) dx √x 1 4 2 ln 6 3 2 11. (x 1) dx 30. 8ex dx − 0 ln 3 2 9 2 12. (x + 1/x) dx 31. 2t dt 1 8 3 e − 3 13. x5 + 2 dx 32. dx 3 e2 x − 1 − 3 14. (x 1)(3x + 2) dx 33. x2 1 dx 1 − 2 | − | − 4 2 15. (√t 2/√t) dt 2 − 34. x x dx 1 1 | − | − 8 1 2 16. √3 r + dr 3 35. (x 2 x ) dx 1 √r 1 − | | − 0 3 2 17. (x + 1) dx 36. (x2 x 1 ) dx 1 − | − | − 0 e x2 + x + 1 2 18. dx x 37. f(x) dx where 1 0 4 9 1 2 x if 0 x< 1, 19. √x + dx f(x)= 5 ≤ √x x , if 1 x 2. 4 ≤ ≤ 1 π 20. √4 x5 + √5 x4 dx 38. f(x) dx where 0 π − 8 x 1 x, if π x 0, 21. − dx f(x)= − ≤ ≤ √3 2 sin x, if 0

39. Compute 40. Compute 2 2 3 I = 2x 1+ x dx 2 n In = 2x 1+ x dx. 0 in two diﬀerent ways: (i) Expand (1+ x2)3, multiply with 2x, 2 and integrate each term. 41. If f ′(x) = x 1/x and f(1) = 1/2 − (ii) Use the substitution u =1+ x2. ﬁnd f(x).

42. Consider 2 x 1 dx. Let f(x)= x 1 so that 0 | − | | − | x 1 if x 1 f(x)= − ≥ 1 x if x< 1 − Deﬁne x2 2 x if x 1 F (x)= −x2 ≥ x if x< 1 − 2 Then since F is an antiderivative of f we have by the Fundamental Theorem of Calculus: 2 2 22 02 x 1 dx = f(x) dx = F (2) F (0) = ( 2) (0 ) = 0 | − | − 2 − − − 2 0 0 But this integral cannot be zero, f(x) is positive except at one point. How can this be?

BASIC SUBSTITUTIONS

Use a substitution to evaluate the following integrals.

2 u du 2 ln 2x 43. 50. dx 1+ u2 x 1 1 2 x dx 44. ln(2x2) 1+ x2 51. dx 1 x π/3 2 45. sin θ cos θ dθ √2 2 10 π/4 52. ξ(1+2ξ ) dξ ξ=0 3 1 46. , dr 3 2 r ln r 53. sin ρ cos 2ρ)4 dρ sin 2x 47. dx 2 1 + cos2 x α2 sin 2x 54. αe− dα 48. dx 1 + sin x 1 1 e t 49. z 1 z2 dz 55. 2 dt 0 − t

REVIEW OF THE INVERSE TRIGONOMETRIC FUNCTIONS

56. Group problem. The inverse sine function is the inverse function to the (restricted) sine function, i.e. when π/2 θ π/2 we have − ≤ ≤ θ = arcsin(y) y = sin θ. ⇐⇒ 21

The inverse sine function is sometimes called Arc Sine function and denoted θ = 1 arcsin(y). We avoid the notation sin− (x) which is used by some as it is ambiguous 1 (it could stand for either arcsin x or for (sin x)− = 1/(sin x)). (i) If y = sin θ, express sin θ, cos θ, and tan θ in terms of y when 0 θ <π/2. ≤ (ii) If y = sin θ, express sin θ, cos θ, and tan θ in terms of y when π/2 <θ π. ≤ (iii) If y = sin θ, express sin θ, cos θ, and tan θ in terms of y when π/2 <θ< 0. − dy (iv) Evaluate using the substitution y = sin θ, but give the final answer in 2 1 y terms of y. − 57. Group problem. Express in simplest form: 1 1 ln 4 (i) cos(arcsin− (x)); (ii) tan arcsin ; (iii) sin 2 arctan a ln 16 58. Group problem. Draw the graph of y = f(x) = arcsin sin(x) , for 2π x +2π. − ≤ ≤ Make sure you get the same answer as your graphing calculator. √3/2 dx 59. Use the change of variables formula to evaluate first using the substi- √1 x2 1/2 tution x = sin u and then using the substitution x = cos u. − 60. The inverse tangent function is the inverse function to the (restricted) tangent function, i.e. for π/2 <θ<π/2 we have θ = arctan(w) w = tan θ. ⇐⇒ The inverse tangent function is sometimes called Arc Tangent function and denoted 1 θ = arctan(y). We avoid the notation tan− (x) which is used by some as it is ambiguous 1 (it could stand for either arctan x or for (tan x)− = 1/(tan x)). (i) If w = tan θ, express sin θ and cos θ in terms of w when (a) 0 θ <π/2 (b) π/2 <θ π (c) π/2 <θ< 0 ≤ ≤ − dw (ii) Evaluate using the substitution w = tan θ, but give the final answer in 1+ w2 terms of w. 61. Use the substitution x = tan(θ) to find the following integrals. Give the final answer in terms of x. (a) 1+ x2 dx 1 (b) dx (1 + x2)2 dx (c) √1+ x2 Evaluate these integrals:

dx x dx √3/2 62. 65. dx 2 4 68. √1 x √1 4x 2 0 √1 x − − − dx 1/2 dx 63. 66. dx 2 2 69. , √4 x 1/2 √4 x x2 + 1 − − − 1 dx dx dx 64. 67. 70. , 2 2 2 2 √2x x 1 √4 x x + a − − − 22

dx dx a√3 dx 71. 2 , 72. 2 74. . 7 + 3x 3x + 6x + 6 x2 + a2 a √3 dx 73. , x2 + 1 1

INTEGRATION BY PARTS AND REDUCTION FORMULAE

75. Evaluate xn ln x dx where n = 1. − 76. Evaluate eax sin bx dx where a2 + b2 = 0. [Hint: Integrate by parts twice.] 77. Evaluate eax cos bx dx where a2 + b2 = 0. 78. Prove the formula n x n x n 1 x x e dx = x e n x − e dx − and use it to evaluate x2ex dx. 79. Prove the formula

n 1 n 1 n 1 n 2 sin x dx = cos x sin − x + − sin − x dx, n = 0 − n n

80. Evaluate sin2 x dx. Show that the answer is the same as the answer you get using the half angle formula. π 81. Evaluate sin14 x dx. 0 82. Prove the formula

n 1 n 1 n 1 n 2 cos x dx = sin x cos − x + − cos − x dx, n = 0 n n π/4 and use it to evaluate cos4 x dx. 0 83. Prove the formula m+1 n m n x (ln x) n m n 1 x (ln x) dx = x (ln x) − dx, m = 1, m + 1 − m + 1 − and use it to evaluate the following integrals:

84. ln x dx 85. (ln x)2 dx 86. x3(ln x)2 dx 1 87. Evaluate x− ln x dx by another method. [Hint: the solution is short!] 88. For an integer n> 1 derive the formula

n 1 n 1 n 2 tan x dx = tan − x tan − x dx n 1 − − 23

π/4 Using this, ﬁnd tan5 x dx by doing just one explicit integration. 0 Use the reduction formula from example 8.4 to compute these integrals: dx 89. (1 + x2)3 dx 90. (1 + x2)4 xdx 91. [Hint: x/(1 + x2)ndx is easy.] (1 + x2)4 1+ x 92. dx (1 + x2)2 93. Group problem. The reduction formula from example 8.4 is valid for all n =0. In particular, n does not have to be an integer, and it does not have to be positive. dx Find a relation between 1+ x2 dx and by setting n = 1 . √1+ x2 − 2 94. Apply integration by parts to 1 dx x 1 1 Let u = x and dv = dx. This gives us, du = −x2 dx and v = x. 1 1 1 dx =( )(x) x − dx x x − x2 Simplifying 1 1 dx =1+ dx x x and subtracting the integral from both sides gives us 0 = 1. How can this be?

INTEGRATION OF RATIONAL FUNCTIONS Express each of the following rational functions as a polynomial plus a proper rational function. (See 9.1 for deﬁnitions.) § x3 x3 x2 x 5 95. , 97. − − − . x3 4 x3 4 − − x3 + 2x x3 1 96. , 98. − . x3 4 x2 1 − −

COMPLETING THE SQUARE Write ax2 +bx+c in the form a(x+p)2 +q, i.e. ﬁnd p and q in terms of a, b, and c (this procedure, which you might remember from high school algebra, is called “completing the square.”). Then evaluate the integrals

dx 99. , 102. Use the method of equating coeﬃ- x2 + 6x + 8 cients to ﬁnd numbers A, B, C such that 2 dx x + 3 A B C 100. , = + + x2 + 6x + 10 x(x + 1)(x 1) x x + 1 x 1 − − and then evaluate the integral dx x2 + 3 101. . dx. 5x2 + 20x + 25 x(x + 1)(x 1) − 24

dx 103. Do the previous problem using the 113. 1+ ex Heaviside trick. x2 + 3 dx 104. Find the integral dx. 114. x2(x 1) x(x2 + 1) − dx Evaluate the following integrals: 115. x(x2 + 1)2 2 4 − x 1 105. 2 − dx dx 5 x + 1 116. − x2(x 1) x3 dx − 106. 1 x4 + 1 117. dx (x 1)(x 2)(x 3) x5 dx − − − 107. x2 + 1 x2 1 118. dx − (x 1)(x 2)(x 3) x5 dx − − − 108. x3 + 1 x4 1 119. dx − (x 1)(x 2)(x 3) x3 − − − 109. dx 120. Group problem. x2 1 − 2 dx e3x dx (a) Compute where h 110. 1 x(x h) e4x 1 is a positive number. − − ex dx (b) What happens to your answer 111. √1+ e2x to (a) when h 0+ ? → ex dx 2 dx 112. (c) Compute . e2x + 2ex + 2 x2 1 MISCELLANEOUS AND MIXED INTEGRALS 121. Find the area of the region bounded by the curves 2 x2 8x + 7 x = 1, x = 2, y = , y = − . x2 4x + 5 x2 8x + 16 − − 122. Let P be the piece of the parabola y = x2 on which 0 x 1. ≤ ≤ (i) Find the area between P, the x-axis and the line x = 1. (ii) Find the length of P. 123. Let a be a positive constant and x F (x)= sin(aθ) cos(θ)dθ. 0 [Hint: use a trig identity for sin A cos B, or wait until we have covered complex exponentials and then come back to do this problem.] (i) Find F (x) if a = 1. (ii) Find F (x) if a = 1. (Don’t divide by zero.) Evaluate the following integrals:

a 4 x dx 124. x sin x dx 126. 2 0 3 √x 1 −

a 1/3 x dx 2 127. 125. x cos x dx 2 1/4 √1 x 0 − 25

4 dx x 128. 138. dx 2 3 3 x√x 1 (x 1) − − 4 x dx 139. dx 129. (x 1)3(x + 1) x2 + 2x + 17 − 1 x4 140. dx 130. dx √1 2x x2 x2 36 − − − e 141. x ln x dx x4 131. dx 1 36 x2 − 142. 2x ln(x + 1) dx (x2 +1)dx 132. x4 x2 e3 − 143. x2 ln x dx (x2 +3)dx e2 133. x4 2x2 e − 144. x(ln x)3 dx 1 x 134. e (x + cos(x))dx 145. arctan(√x)dx 135. (ex + ln(x)) dx 146. x(cos x)2 dx 3x2 + 2x 2 π 136. − dx x3 1 147. 1 + cos(6w)dw − 0 x4 1 137. dx 148. dx x4 16 1 + sin(x) −

149. Find dx x(x 1)(x 2)(x 3) − − − and (x3 +1)dx x(x 1)(x 2)(x 3) − − − 150. Find dx x3 + x2 + x + 1 151. Group problem. You don’t always have to find the antiderivative to find a definite integral. This problem gives you two examples of how you can avoid finding the antiderivative. (i) To find π/2 sin x dx I = sin x + cos x 0 you use the substitution u = π/2 x. The new integral you get must of course be equal − to the integral I you started with, so if you add the old and new integrals you get 2I. If you actually do this you will see that the sum of the old and new integrals is very easy to compute. π/2 (ii) Use the same trick to find sin2 x dx 0 2 2 2 152. Group problem. Graph the equation x 3 + y 3 = a 3 . Compute the area bounded by this curve. 26

153. Group problem. The Bow-Tie Graph. Graph the equation y2 = x4 x6. Compute − the area bounded by this curve. 154. Group problem. The Fan-Tailed Fish. Graph the equation 1 x y2 = x2 − . 1+ x Find the area enclosed by the loop. (Hint: Rationalize the denominator of the integrand.) 155. Find the area of the region bounded by the curves x x = 2, y = 0, y = x ln 2 156. Find the volume of the solid of revolution obtained by rotating around the x axis the − region bounded by the lines x = 5, x = 10, y = 0, and the curve x y = . √x2 + 25

157. 1 How to ﬁnd the integral of f(x)= cos x (i) Verify the answer given in the table in the lecture notes. (ii) Note that 1 cos x cos x = = , cos x cos2 x 1 sin2 x and apply the substitution s = sin x followed by− a partial fraction decomposition to dx compute cos x . RATIONALIZING SUBSTITUTIONS Recall that a rational function is the ratio of two polynomials. 158. Prove that the family of rational functions is closed under taking sums, products, quotients (except do not divide by the zero polynomial), and compositions.

To integrate rational functions of x and √1+ x2 one may do a trigonometric substitution, e.g., x = tan(θ) and 1+tan2(θ) = sec2(θ). This turns the problem into a trig integral. Or one could use 1 + sinh2(t) = cosh2(t) and convert the problem into a rational function of et. Another technique which works is to use the parameterization of the hyperbola by rational functions:

1 1 1 1 x = (t ) y = (t + ) 2 − t 2 t 159. Show that y2 x2 = 1 and hence y = √1+ x2. − Use this to rationalize the integrals, i.e, make them into an integral of a rational function of t. You do not need to integrate the rational function.

ds 160. 1+ x2 dx 162. √s2 + 2s + 3 x4 161. dx √1+ x2

163. Show that t = x + y = x + √1+ x2. 27

Hence if g(x) dx = f(t) dt = F (t)+ C then g(x) dx = F (x + 1 x2)+ C. −

164. Note that x = y2 1. Show that t is a function of y. − Express these integrals as integrals of rational functions of t.

dy s4 165. 167. ds (y2 1)1/2 (s2 36)3/2 − − y4 ds 166. dy 168. (y2 1)1/2 (s2 + 2s)1/2 −

x 2 1 2 169. Note that 1 = ( y ) +( y ) . What substitution would rationalize integrands which have √1 z2 in them? Show how to write t as a function of z. − Express these integrals as integrals of rational functions of t.

s4 170. 1 z2 dz 173. ds − (36 s2)3/2 − dz 171. √1 z2 − z2 ds 172. dz 174. √1 z2 (s + 5)√s2 + 5s −

Integrating a rational function of sin and cos r(sin(θ), cos(θ)) dθ Examples of such integrals are: (cos θ)2 (cos θ)(sin θ) + 1 − dθ (cos θ)2 + (sin θ)3 + (cos θ) + 1 or (sin θ)3(cos θ) + (cos θ) + (sin θ) + 1 dθ (cos θ)2(sin θ)3 (cos θ) − The goal of the following problems is to show that such integrals can be rationalized, not to integrat the rational function.

175. Substitute z = sin(θ) and express r(sin(θ), cos(θ)) dθ as a rational function of z and √1 z2. − 176. Express it as rational function of t.

177. Express t as a function of θ.

π Is 2 a rational number? 28

178. Consider the integral: 1 x2 dx − Substitute u = 1 x2 so − u = 1 x2 − 1 x = √1 u = (1 u) 2 − − 1 1 dx =( )(1 u)− 2 ( 1) du 2 − − Hence 1 1 1 x2 dx = √u ( )(1 u)− 2 ( 1) du − 2 − − Take the deﬁnite integral from x = 1 to x = 1 and note that u = 0 when x = 1 − − and u = 0 also when x = 1, so 1 0 1 1 1 x2 dx = √u( )(1 u)− 2 ( 1) du = 0 1 − 0 2 − − − 0 The last being zero since 0 anything is always 0. But the integral on the left is equal to half the area of the unit disk, hence π . Therefor π = 0 which is a rational number. 2 2 How can this be? 29

Chapter 2: Taylor’s Formula and Inﬁnite Series

All continuous functions which vanish at x = a are approximately equal at x = a, but some are more approximately equal than others.

11. Taylor Polynomials

Suppose you need to do some computation with a complicated function y = f(x), and suppose that the only values of x you care about are close to some constant x = a. Since polynomials are simpler than most other functions, you could then look for a polynomial y = P (x) which somehow “matches” your function y = f(x) for values of x close to a. And you could then replace your function f with the polynomial P , hoping that the error you make isn’t too big. Which polynomial you will choose depends on when you think a polynomial “matches” a function. In this chapter we will say that a polynomial P of degree n matches a function f at x = a if P has the same value and the same derivatives of order 1, 2,..., n at x = a as the function f. The polynomial which matches a given function at some point x = a is the Taylor polynomial of f. It is given by the following formula.

11.1. Deﬁnition. The Taylor polynomial of a function y = f(x) of degree n at a point a is the polynomial

(n) a f ′′(a) 2 f (a) n (6) T f(x)= f(a)+ f ′(a)(x a)+ (x a) + + (x a) . n − 2! − n! − (Recall that n! = 1 2 3 n, and by deﬁnition 0! = 1.

11.2. Theorem. The Taylor polynomial has the following property: it is the only polynomial P (x) of degree n whose value and whose derivatives of orders 1, 2, . . . , and n are the same as those of f, i.e. it’s the only polynomial of degree n for which

(n) (n) P (a)= f(a), P ′(a)= f ′(a), P ′′(a)= f ′′(a), ..., P (a)= f (a) holds.

Proof. We do the case a = 0, for simplicity. Let n be given, consider a polynomial P (x) of degree n, say,

2 3 n P (x)= a0 + a1x + a2x + a3x + + anx , 30 and let’s see what its derivatives look like. They are: 2 3 4 P (x) = a0 + a1x + a2x + a3x + a4x + 2 3 P ′(x) = a1 + 2a2x + 3a3x + 4a4x + (2) 2 P (x) = 1 2a2 + 2 3a3x + 3 4a4x + (3) P (x) = 1 2 3a3 + 2 3 4a4x + (4) P (x) = 1 2 3 4a4 + When you set x = 0 all the terms which have a positive power of x vanish, and you are left with the ﬁrst entry on each line, i.e.

(2) (3) P (0) = a0, P ′(0) = a1, P (0) = 2a2, P (0) = 2 3a3, etc. and in general (k) P (0) = k!ak for 0 k n. ≤ ≤ For k n + 1 the derivatives p(k)(x) all vanish of course, since P (x) is a polynomial of ≥ degree n. Therefore, if we want P to have the same values and derivatives at x = 0 of orders (k) (k) 1,,. . . , n as the function f, then we must have k!ak = P (0) = f (0) for all k n. ≤ Thus f (k)(0) ak = for 0 k n. k! ≤ ≤

12. Examples

Note that the zeroth order Taylor polynomial is just a constant, a T0 f(x)= f(a), while the ﬁrst order Taylor polynomial is a T f(x)= f(a)+ f ′(a)(x a). 1 − This is exactly the linear approximation of f(x) for x close to a which was derived in 1st semester calculus. The Taylor polynomial generalizes this ﬁrst order approximation by providing “higher order approximations” to f.

y = f(x) y = f(x) y = f(x)

y = T1f(x)

y = T0f(x) y = T2f(x)

Figure 1. The Taylor polynomials of degree 0, 1 and 2 of f(x)= ex at a = 0. The zeroth order Taylor polynomial has the right value at x = 0 but it doesn’t know whether or not the function f is increasing at x = 0. The ﬁrst order Taylor polynomial has the right slope at x = 0, but it doesn’t see if the graph of f is curved up or down at x = 0. The second order Taylor polynomial also has the right curvature at x = 0. 31

a Most of the time we will take a = 0 in which case we write Tnf(x) instead of Tn f(x), and we get a slightly simpler formula

(n) f ′′(0) 2 f (0) n (7) Tnf(x)= f(0) + f ′(0)x + x + + x . 2! n!

You will see below that for many functions f(x) the Taylor polynomials Tnf(x) give better and better approximations as you add more terms (i.e. as you increase n). For this reason the limit when n is often considered, which leads to the inﬁnite sum →∞ f ′′(0) 2 f ′′′(0) 3 T f(x)= f(0) + f ′(0)x + x + x + ∞ 2! 3! At this point we will not try to make sense of the “sum of inﬁnitely many numbers”.

12.1. Example: Compute the Taylor polynomials of degree 0, 1 and 2 of f(x)= ex at a = 0, and plot them. One has x x x f(x)= e = f ′(x)= e = f ′′(x)= e , ⇒ ⇒ so that f(0) = 1, f ′(0) = 1, f ′′(0) = 1. Therefore the ﬁrst three Taylor polynomials of ex at a = 0 are

y =1+ x + x2

3 2 y =1+ x + 2 x

1 2 y =1+ x + 2 x

x y = e

=1+ y y =1+ x 1 x2 − 2 Figure 2. The top edge of the shaded region is the graph of y = ex. The graphs are of the functions y =1+ x + Cx2 for various values of C. These graphs all are tangent at x = 0, but one of the parabolas matches the graph of y = ex better than any of the others. 32

T0f(x) = 1

T1f(x)=1+ x 1 T f(x)=1+ x + x2. 2 2 The graphs are found in Figure 2. As you can see from the graphs, the Taylor polynomial x x T0f(x) of degree 0 is close to e for small x, by virtue of the continuity of e x The Taylor polynomial of degree 0, i.e. T0f(x) = 1 captures the fact that e by virtue of its continuity does not change very much if x stays close to x = 0.

The Taylor polynomial of degree 1, i.e. T1f(x)=1+ x corresponds to the tangent line to the graph of f(x) = ex, and so it also captures the fact that the function f(x) is increasing near x = 0. x Clearly T1f(x) is a better approximation to e than T0f(x).

The graphs of both y = T0f(x) and y = T1f(x) are straight lines, while the graph of y = ex is curved (in fact, convex). The second order Taylor polynomial captures this convexity. In fact, the graph of y = T2f(x) is a parabola, and since it has the same ﬁrst and second derivative at x = 0, its curvature is the same as the curvature of the graph of y = ex at x = 0. 2 x So it seems that y = T2f(x)=1+ x + x /2 is an approximation to y = e which beats both T0f(x) and T1f(x).

12.2. Example: Find the Taylor polynomials of f(x) = sin x. When you start computing the derivatives of sin x you ﬁnd (3) f(x) = sin x, f ′(x) = cos x, f ′′(x)= sin x, f (x)= cos x, − − and thus f (4)(x) = sin x. So after four derivatives you’re back to where you started, and the sequence of derivatives of sin x cycles through the pattern sin x, cos x, sin x, cos x, sin x, cos x, sin x, cos x, sin x, ... − − − − on and on. At x = 0 you then get the following values for the derivatives f (j)(0), j 1 2 3 4 5 6 7 8 f (j)(0) 0 1 0 1 0 1 0 1 − − This gives the following Taylor polynomials

T0f(x) = 0

T1f(x)= x

T2f(x)= x x3 T3f(x)= x − 3! x3 T4f(x)= x − 3! x3 x5 T5f(x)= x + − 3! 5! (2) Note that since f (0) = 0 the Taylor polynomials T1f(x) and T2f(x) are the same! The second order Taylor polynomial in this example is really only a polynomial of degree 1. In general the Taylor polynomial Tnf(x) of any function is a polynomial of degree at most n, and this example shows that the degree can sometimes be strictly less. 33 T1f(x) T5f(x) T9f(x)

y = sin x

π 2π 2π π − −

T3f(x) T7f(x) T11f(x)

Figure 3. Taylor polynomials of f(x) = sin x

12.3. Example – Compute the Taylor polynomials of degree two and three of f(x)=1+ x + x2 + x3 at a = 3. Solution: Remember that our notation for the nth a degree Taylor polynomial of a function f at a is Tn f(x), and that it is deﬁned by (6).

We have

2 f ′(x)=1+2x + 3x , f ′′(x)=2+6x, f ′′′(x) = 6

Therefore f(3) = 40, f ′(3) = 34, f ′′(3) = 20, f ′′′(3) = 6, and thus

20 (8) T sf(x) = 40 + 34(x 3) + (x 3)2 = 40 + 34(x 3) + 10(x 3)2. 2 − 2! − − −

Why don’t we expand the answer? You could do this (i.e. replace (x 3)2 by x2 6x + 9 − − throughout and sort the powers of x), but as we will see in this chapter, the Taylor a polynomial Tn f(x) is used as an approximation for f(x) when x is close to a. In this example T 3f(x) is to be used when x is close to 3. If x 3 is a small number then the 2 − successive powers x 3, (x 3)2, (x 3)3, . . . decrease rapidly, and so the terms in (8) − − − are arranged in decreasing order.

We can also compute the third degree Taylor polynomial. It is

20 6 T 3f(x) = 40 + 34(x 3) + (x 3)2 + (x 3)3 3 − 2! − 3! − = 40 + 34(x 3) + 10(x 3)2 +(x 3)3. − − −

If you expand this (this takes a little work) you ﬁnd that

40 + 34(x 3) + 10(x 3)2 +(x 3)3 =1+ x + x2 + x3. − − −

So the third degree Taylor polynomial is the function f itself! Why is this so? Because of Theorem 11.2! Both sides in the above equation are third degree polynomials, and their derivatives of order 0, 1, 2 and 3 are the same at x = 3, so they must be the same polynomial. 34

13. Some special Taylor polynomials

Here is a list of functions whose Taylor polynomials are suﬃciently regular that you can write a formula for the nth term. 2 3 n x x x x Tne =1+ x + + + + 2! 3! n! 3 5 7 2n+1 x x x n x T2n+1 sin x = x + + +( 1) { } − 3! 5! − 7! − (2n + 1)! 2 4 6 2n x x x n x T2n cos x = 1 + + +( 1) { } − 2! 4! − 6! − (2n)!

1 2 3 4 n Tn =1+ x + x + x + x + + x (Geometric Series) 1 x − 2 3 4 n x x x n+1 x Tn ln(1 + x) = x + + +( 1) { } − 2 3 − 4 − n All of these Taylor polynomials can be computed directly from the definition, by repeatedly differentiating f(x). Another function whose Taylor polynomial you should know is f(x)=(1+x)a, where a is a constant. You can compute Tnf(x) directly from the definition, and when you do this you find

a a(a 1) 2 a(a 1)(a 2) 3 (9) Tn (1 + x) =1+ ax + − x + − − x { } 1 2 1 2 3 a(a 1) (a n + 1) + + − − xn. 1 2 n This formula is called Newton’s binomial formula. The coeﬃcient of xn is called a binomial coeﬃcient, and it is written

a a(a 1) (a n + 1) (10) = − − . n n!

a When a is an integer n is also called “a choose n.” Note that you already knew special cases of the binomial formula: when a is a positive integer the binomial coeﬃcients are just the numbers in Pascal’s triangle. When a = 1 − the binomial formula is the Geometric series.

14. The Remainder Term

The Taylor polynomial Tnf(x) is almost never exactly equal to f(x), but often it is a good approximation, especially if x is small. To see how good the approximation is we deﬁne the “error term” or, “remainder term”.

14.1. Deﬁnition. If f is an n times diﬀerentiable function on some interval containing a, then Ra f(x)= f(x) T af(x) n − n th is called the n order remainder (or error) term in the Taylor polynomial of f. If a = 0, as will be the case in most examples we do, then we write

Rnf(x)= f(x) Tnf(x). − 35

1 3 14.2. Example. If f(x) = sin x then we have found that T3f(x)= x x , so that − 6 1 3 R3 sin x = sin x x + x . { } − 6 This is a completely correct formula for the remainder term, but it’s rather useless: there’s 1 3 nothing about this expression that suggests that x 6 x is a much better approximation 1 3 − to sin x than, say, x + 6 x . The usual situation is that there is no simple formula for the remainder term.

14.3. An unusual example, in which there is a simple formula for Rnf(x). Consider f(x) = 1 x + 3 x2 15 x3. − − Then you ﬁnd

2 3 T2f(x) = 1 x + 3 x , so that R2f(x)= f(x) T2f(x)= 15 x . − − − th The moral of this example is this: Given a polynomial f(x) you ﬁnd its n degree Taylor polynomial by taking all terms of degree n in f(x); the remainder Rnf(x) then consists ≤ of the remaining terms.

14.4. Another unusual, but important example where you can compute Rnf(x). Consider the function 1 f(x)= . 1 x − Then repeated diﬀerentiation gives

1 (2) 1 2 (3) 1 2 3 f ′(x)= , f (x)= , f (x)= , ... (1 x)2 (1 x)3 (1 x)4 − − − and thus 1 2 3 n f (n)(x)= . (1 x)n+1 − Consequently, 1 f (n)(0) = n! = f (n)(0) = 1, ⇒ n! and you see that the Taylor polynomials of this function are really simple, namely

2 3 4 n Tnf(x)=1+ x + x + x + x + + x . But this sum should be really familiar: it is just the Geometric Sum (each term is x times the previous term). Its sum is given by5

n+1 2 3 4 n 1 x Tnf(x)=1+ x + x + x + x + + x = − , 1 x − which we can rewrite as 1 xn+1 xn+1 Tnf(x)= = f(x) . 1 x − 1 x − 1 x − − − The remainder term therefore is xn+1 Rnf(x)= f(x) Tnf(x)= . − 1 x −

5Multiply both sides with 1 x to verify this, in case you had forgotten the formula! − 36

15. Lagrange’s Formula for the Remainder Term

15.1. Theorem. Let f be an n + 1 times diﬀerentiable function on some interval I containing x = 0. Then for every x in the interval I there is a ξ between 0 and x such that f (n+1)(ξ) R f(x)= xn+1. n (n + 1)! (ξ between 0 and x means either 0 <ξ

15.2. Estimate of remainder term. If f is an n + 1 times diﬀerentiable function on an interval containing x = 0, and if you have a constant M such that

( ) f (n+1)(t) M for all t between 0 and x, † ≤ then M x n+1 Rnf(x) | | . | |≤ (n + 1)!

Proof. We don’t know what ξ is in Lagrange’s formula, but it doesn’t matter, for wherever it is, it must lie between 0 and x so that our assumption ( ) implies f (n+1)(ξ) † | |≤ M. Put that in Lagrange’s formula and you get the stated inequality.

15.3. How to compute e in a few decimal places. Consider f(x) = ex. We computed the Taylor polynomials before. If you set x = 1, then you get e = f(1) = Tnf(1) + Rnf(1), and thus, taking n = 8, 1 1 1 1 1 1 1 1 e =1+ + + + + + + + + R (1). 1! 2! 3! 4! 5! 6! 7! 8! 8 By Lagrange’s formula there is a ξ between 0 and 1 such that f (9)(ξ) eξ R (1) = 19 = . 8 9! 9! (remember: f(x)= ex, so all its derivatives are also ex.) We don’t really know where ξ is, ξ but since it lies between 0 and 1 we know that 1

15.4. Error in the approximation sin x x. In many calculations involving sin x ≈ for small values of x one makes the simplifying approximation sin x x, justiﬁed by the ≈ known limit sin x lim = 1. x 0 → x Question: How big is the error in this approximation? To answer this question, we use Lagrange’s formula for the remainder term again. Let f(x) = sin x. Then the ﬁrst degree Taylor polynomial of f is

T1f(x)= x. The approximation sin x x is therefore exactly what you get if you approximate f(x)= ≈ sin x by its ﬁrst degree Taylor polynomial. Lagrange tells us that

f(x)= T1f(x)+ R1f(x), i.e. sin x = x + R1f(x), where, since f ′′(x)= sin x, − f ′′(ξ) 2 1 2 R1f(x)= x = sin ξ x 2! − 2 for some ξ between 0 and x. As always with Lagrange’s remainder term, we don’t know where ξ is precisely, so we have to estimate the remainder term. The easiest way to do this (but not the best: see below) is to say that no matter what ξ is, sin ξ will always be between 1 and 1. Hence − the remainder term is bounded by 1 2 ( ) R1f(x) x , ¶ | |≤ 2 and we find that x 1 x2 sin x x + 1 x2. − 2 ≤ ≤ 2 Question: How small must we choose x to be sure that the approximation sin x x isn’t ≈ off by more than 1% ? 1 2 If we want the error to be less than 1% of the estimate, then we should require 2 x to be less than 1% of x , i.e. | | 1 x2 < 0.01 x x < 0.02 2 | | ⇔ | | So we have shown that, if you choose x < 0.02, then the error you make in approximating | | sin x by just x is no more than 1%. A final comment about this example: the estimate for the error we got here can be improved quite a bit in two different ways: (1) You could notice that one has sin x x for all x, so if ξ is between 0 and x, then | |≤ sin ξ ξ x , which gives you the estimate | | ≤ | | ≤ | | 1 3 1 2 R1f(x) x instead of x as in ( ). | |≤ 2 | | 2 ¶

(2) For this particular function the two Taylor polynomials T1f(x) and T2f(x) are the same (because f ′′(0) = 0). So T2f(x)= x, and we can write

sin x = f(x)= x + R2f(x), In other words, the error in the approximation sin x x is also given by the second order ≈ remainder term, which according to Lagrange is given by

16. The limit as x 0, keeping n ﬁxed → 16.1. Little-oh. Lagrange’s formula for the remainder term lets us write a function y = f(x), which is deﬁned on some interval containing x = 0, in the following way

(2) (n) (n+1) f (0) 2 f (0) n f (ξ) n+1 (11) f(x)= f(0) + f ′(0)x + x + + x + x 2! n! (n + 1)! The last term contains the ξ from Lagrange’s theorem, which depends on x, and of which you only know that it lies between 0 and x. For many purposes it is not necessary to know the last term in this much detail – often it is enough to know that “in some sense” the last term is the smallest term, in particular, as x 0 it is much smaller than x, or 2 n → x , or, . . . , or x :

16.2. Theorem. If the n + 1st derivative f (n+1)(x) is continuous at x = 0 then the (n+1) n+1 remainder term Rnf(x)= f (ξ)x /(n + 1)! satisﬁes R f(x) lim n = 0 x 0 k → x for any k = 0, 1, 2,...,n.

(n+1) (n+1) Proof. Since ξ lies between 0 and x, one has limx 0 f (ξ) = f (0), and → therefore n+1 Rnf(x) (n+1) x (n+1) n+1 k (n+1) lim = lim f (ξ) = lim f (ξ) x − = f (0) 0 = 0. x 0 k x 0 k x 0 → x → x →

So we can rephrase (11) by saying

(2) (n) f (0) 2 f (0) n f(x)= f(0) + f ′(0)x + x + + x + remainder 2! n! n n 1 2 where the remainder is much smaller than x , x − , ..., x , x or 1. In order to express the condition that some function is “much smaller than xn,” at least for very small x, Landau introduced the following notation which many people ﬁnd useful.

16.3. Deﬁnition. “o(xn)” is an abbreviation for any function h(x) which satisﬁes h(x) lim = 0. x 0 n → x So you can rewrite (11) as

(2) (n) f (0) 2 f (0) n n f(x)= f(0) + f ′(0)x + x + + x + o(x ). 2! n! The nice thing about Landau’s little-oh is that you can compute with it, as long as you obey the following (at ﬁrst sight rather strange) rules which will be proved in class xn o(xm)= o(xn+m) o(xn) o(xm)= o(xn+m) xm = o(xn) if n

x x2 x3 x4 10 x x20

Figure 4. How the powers stack up. All graphs of y = xn (n > 1) are tangent to the x-axis at the origin. But the larger the exponent n the “ﬂatter” the graph of y = xn is.

16.4. Example: prove one of these little-oh rules. Let’s do the first one, i.e. let’s show that xn o(xm) is o(xn+m) as x 0. → Remember, if someone writes xn o(xm), then the o(xm) is an abbreviation for some m n m function h(x) which satisfies limx 0 h(x)/x = 0. So the x o(x ) we are given here n → really is an abbreviation for x h(x). We then have xnh(x) h(x) lim = lim = 0, since h(x)= o(xm). x 0 n+m x 0 m → x → x 16.5. Can you see that x3 = o(x2) by looking at the graphs of these functions? A picture is of course never a proof, but have a look at figure 4 which shows you the graphs of y = x, x2, x3, x4, x5 and x10. As you see, when x approaches 0, the graphs of higher powers of x approach the x-axis (much?) faster than do the graphs of lower powers. You should also have a look at figure 5 which exhibits the graphs of y = x2, as well 1 1 1 as several linear functions y = Cx (with C = 1, 2 , 5 and 10 .) For each of these linear functions one has x2 < Cx if x is small enough; how small is actually small enough depends on C. The smaller the constant C, the closer you have to keep x to 0 to be sure that x2 is smaller than Cx. Nevertheless, no matter how small C is, the parabola will eventually always reach the region below the line y = Cx.

16.6. Example: Little-oh arithmetic is a little funny. Both x2 and x3 are functions which are o(x), i.e. x2 = o(x) and x3 = o(x) Nevertheless x2 = x3. So in working with little-oh we are giving up on the principle that says that two things which both equal a third object must themselves be equal; in other words, a = b and b = c implies a = c, but not when you’re using little-ohs! You can also put it like this: just because two quantities both are much smaller than x, they don’t have to be equal. In particular, you can never cancel little-ohs!!! 40 y = x

y = x2

y = x/2

y = x/5 y = x/10 y = x/20

Figure 5. x2 is smaller than any multiple of x, if x is small enough. Compare the quadratic function y = x2 with a linear function y = Cx. Their graphs are a parabola and a straight line. Parts of the parabola may lie above the line, but as x 0 the parabola will always duck underneath ց the line.

In other words, the following is pretty wrong

o(x2) o(x2) = 0. − 2 2 Why? The two o(x )’s both refer to functions h(x) which satisfy limx 0 h(x)/x = 0, but → there are many such functions, and the two o(x2)’s could be abbreviations for diﬀerent functions h(x). Contrast this with the following computation, which at ﬁrst sight looks wrong even though it is actually right: o(x2) o(x2)= o(x2). − In words: if you subtract two quantities both of which are negligible compared to x2 for small x then the result will also be negligible compared to x2 for small x.

16.7. Computations with Taylor polynomials. The following theorem is very useful because it lets you compute Taylor polynomials of a function without diﬀerentiating it.

16.8. Theorem. If f(x) and g(x) are n + 1 times diﬀerentiable functions then

n (12) Tnf(x)= Tng(x) f(x)= g(x)+ o(x ). ⇐⇒ In other words, if two functions have the same nth degree Taylor polynomial, then their diﬀerence is much smaller than xn, at least, if x is small.

In principle the deﬁnition of Tnf(x) lets you compute as many terms of the Taylor polynomial as you want, but in many (most) examples the computations quickly get out of hand. To see what can happen go though the following example: 41

16.9. How NOT to compute the Taylor polynomial of degree 12 of f(x)= 1/(1 + x2). Diligently computing derivatives one by one you ﬁnd 1 f(x)= so f(0) = 1 1+ x2 2x f ′(x)= − so f ′(0) = 0 (1 + x2)2 6x2 2 f ′′(x)= − so f ′′(0) = 2 (1 + x2)3 − x x3 f (3)(x) = 24 − so f (3)(0) = 0 (1 + x2)4 1 10x2 + 5x4 f (4)(x) = 24 − so f (4)(0) = 24 = 4! (1 + x2)5 3x + 10x3 3x5 f (5)(x) = 240 − − so f (4)(0) = 0 (1 + x2)6 1 + 21x2 35x4 + 7x6 f (6)(x)= 720 − − so f (4)(0) = 720 = 6! − (1 + x2)7 . .

I’m getting tired of diﬀerentiating – can you ﬁnd f (12)(x)? After a lot of work we give up at the sixth derivative, and all we have found is

1 2 4 6 T6 = 1 x + x x . 1+ x2 − − By the way, 1 78 x2 + 715 x4 1716 x6 + 1287 x8 286 x10 + 13 x12 f (12)(x) = 479001600 − − − (1 + x2)13 and 479001600 = 12!.

16.10. The right approach to ﬁnding the Taylor polynomial of any degree of f(x) = 1/(1 + x2). Start with the Geometric Series: if g(t) = 1/(1 t) then − g(t)=1+ t + t2 + t3 + t4 + + tn + o(tn). Now substitute t = x2 in this limit, − n g( x2) = 1 x2 + x4 x6 + +( 1)nx2n + o x2 − − − − − n Since o x2 = o(x2n) and − 1 1 g( x2)= = , − 1 ( x2) 1+ x2 − − we have found 1 = 1 x2 + x4 x6 + +( 1)nx2n + o(x2n) 1+ x2 − − − By Theorem (16.8) this implies

1 2 4 6 n 2n T2n = 1 x + x x + +( 1) x . 1+ x2 − − − 42

16.11. Example of multiplication of Taylor series. Finding the Taylor series of e2x/(1 + x) directly from the deﬁnition is another recipe for headaches. Instead, you should exploit your knowledge of the Taylor series of both factors e2x and 1/(1 + x):

22x2 23x3 24x4 e2x =1+2x + + + + o(x4) 2! 3! 4! 4 2 =1+2x + 2x2 + x3 + x4 + o(x4) 3 3 1 = 1 x + x2 x3 + x4 + o(x4). 1+ x − − Then multiply these two

1 4 2 e2x = 1 + 2x + 2x2 + x3 + x4 + o(x4) 1 x + x2 x3 + x4 + o(x4) 1+ x 3 3 − − = 1 x + x2 x3 + x4 + o(x4) − − + 2x 2x2 + 2x3 2x4 + o(x4) − 2 3 − 4 4 + 2x 2x + 2x + o(x ) − 4 3 4 4 4 + 3 x 3 x + o(x ) − 2 4 4 + 3 x + o(x ) 1 1 =1+ x + x2 + x3 + x4 + o(x4) (x 0) 3 3 →

16.12. Taylor’s formula and Fibonacci numbers. The Fibonacci numbers are defined as follows: the first two are f0 = 1 and f1 = 1, and the others are defined by the equation

(Fib) fn = fn 1 + fn 2 − − So

f2 = f1 + f0 =1+1=2,

f3 = f2 + f1 =2+1=3,

f4 = f3 + f2 =3+2=5, etc.

The equation (Fib) lets you compute the whole sequence of numbers, one by one, when you are given only the ﬁrst few numbers of the sequence (f0 and f1 in this case). Such an equation for the elements of a sequence is called a recursion relation.

Now consider the function 1 f(x)= . 1 x x2 − − Let 2 3 T f(x)= c0 + c1x + c2x + c3x + ∞ be its Taylor series.

Due to Lagrange’s remainder theorem you have, for any n,

1 2 3 n n = c0 + c1x + c2x + c3x + + cnx + o(x ) (x 0). 1 x x2 → − − 43

Multiply both sides with 1 x x2 and you get − − 2 2 n 1=(1 x x ) (c0 + c1x + c2x + + cn + o(x )) (x 0) − − → 2 n n = c0 + c1x + c2x + + cnx + o(x ) 2 n n c0x c1x cn 1x + o(x ) − − 2 − − − n n c0x cn 2x o(x ) (x 0) − − − − − → 2 3 = c0 +(c1 c0)x +(c2 c1 c0)x +(c3 c2 c1)x + − − − − − n n +(cn cn 1 cn 2)x + o(x ) (x 0) − − − − → Compare the coeﬃcients of powers xk on both sides for k = 0, 1,...,n and you ﬁnd

c0 = 1, c1 c0 =0 = c1 = c0 = 1, c2 c1 c0 =0 = c2 = c1 + c0 = 2 − ⇒ − − ⇒ and in general

cn cn 1 cn 2 =0 = cn = cn 1 + cn 2 − − − − ⇒ − − Therefore the coeﬃcients of the Taylor series T f(x) are exactly the Fibonacci numbers: ∞ cn = fn for n = 0, 1, 2, 3,... Since it is much easier to compute the Fibonacci numbers one by one than it is to compute the derivatives of f(x) = 1/(1 x x2), this is a better way to compute the Taylor series − − of f(x) than just directly from the deﬁnition.

16.13. More about the Fibonacci numbers. In this example you’ll see a trick that lets you compute the Taylor series of any rational function. You already know the trick: find the partial fraction decomposition of the given rational function. Ignoring the case that you have quadratic expressions in the denominator, this lets you represent your rational function as a sum of terms of the form A . (x a)p − These are easy to differentiate any number of times, and thus they allow you to write their Taylor series. Let’s apply this to the function f(x) = 1/(1 x x2) from the example 16.12. First − − we factor the denominator. 1 √5 1 x x2 = 0 x2 + x 1 = 0 x = − ± . − − ⇐⇒ − ⇐⇒ 2 The number 1+ √5 φ = 1.618 033 988 749 89 . . . 2 ≈ is called the Golden Ratio. It satisfies6 1 φ + = √5. φ

The roots of our polynomial x2 + x 1 are therefore − 1 √5 1+ √5 1 x = − − = φ, x+ = − = . − 2 − 2 φ and we can factor 1 x x2 as follows − − 2 2 1 1 x x = (x + x 1) = (x x )(x x+)= (x )(x + φ). − − − − − − − − − − φ

1 2 2 1 √5 1+ √5 6To prove this, use = = − = − . φ 1+ √5 1+ √5 1 √5 2 − 44

So f(x) can be written as 1 1 A B f(x)= = − = + 1 x x2 (x 1 )(x + φ) x 1 x + φ − − − φ − φ The Heaviside trick will tell you what A and B are, namely, 1 1 1 1 A = 1− = − , B = 1 = φ + φ √5 φ + φ √5 The nth derivative of f(x) is n n (n) A( 1) n! B( 1) n! f (x)= − n+1 + − n+1 x 1 (x + φ) − φ Setting x = 0 and dividing by n! ﬁnally gives you the coeﬃcient of xn in the Taylor series of f(x). The result is the following formula for the nth Fibonacci number

(n) n n n+1 f (0) 1 A( 1) n! 1 B( 1) n! n+1 1 cn = = − n+1 + −n+1 = Aφ B n! n! 1 n! − − φ (φ) − φ Using the values for A and B you ﬁnd

1 n+1 1 (13) fn = cn = φ √ − φn+1 5

16.14. Diﬀerentiating Taylor polynomials. If 2 n Tnf(x)= a0 + a1x + a2x + + anx is the Taylor polynomial of a function y = f(x), then what is the Taylor polynomial of its derivative f ′(x)?

16.15. Theorem. The Taylor polynomial of degree n 1 of f ′(x) is given by − n 1 Tn 1 f ′(x) = a1 + 2a2x + + nanx − . − { } In other words, “the Taylor polynomial of the derivative is the derivative of the Taylor polynomial.”

(k) (k+1) Proof. Let g(x)= f ′(x). Then g (0) = f (0), so that 2 n 1 (2) x (n 1) x − Tn 1g(x)= g(0) + g′(0)x + g (0) + + g − (0) − 2! (n 1)! 2 −n 1 (2) (3) x (n) x − ($) = f ′(0) + f (0)x + f (0) + + f (0) 2! (n 1)! − n (k) On the other hand, if Tnf(x)= a0 + a1x + + anx , then ak = f (0)/k!, so that k f (k)(0) ka = f (k)(0) = . k k! (k 1)! − In other words, (3) (2) f (0) 1 a1 = f ′(0), 2a2 = f (0), 3a3 = , etc. 2! So, continuing from ($) you ﬁnd that n 1 Tn 1 f ′(x) = Tn 1g(x)= a1 + 2a2x + + nanx − − { } − as claimed. 45

16.16. Example. We compute the Taylor polynomial of f(x) = 1/(1 x)2 by noting − that 1 f(x)= F ′(x), where F (x)= . 1 x Since − 2 3 n+1 Tn+1F (x)=1+ x + x + x + + x , theorem 16.15 implies that

1 2 3 n Tn =1+2x + 3x + 4x + +(n + 1)x (1 x)2 − 16.17. Example. [Example: Taylor polynomials of arctan x. ] Let f(x) = arctan x. Then know that 1 f ′(x)= . 1+ x2 By substitution of t = x2 in the Taylor polynomial of 1/(1 t) we had found − − 1 2 4 6 n 2n T2n f ′(x) = T2n = 1 x + x x + +( 1) x . { } 1+ x2 − − − This Taylor polynomial must be the derivative of T2n+1f(x), so we have 3 5 2n+1 x x n x T2n+1 arctan x = x + + +( 1) . { } − 3 5 − 2n + 1

17. The limit n , keeping x ﬁxed →∞ 17.1. Sequences and their limits. We shall call a sequence any ordered sequence of numbers a1,a2,a3,...: for each positive integer n we have to specify a number an.

17.2. Examples of sequences. deﬁnition ﬁrst few number in the sequence

↓ ↓ an = n 1, 2, 3, 4,... bn = 0 0, 0, 0, 0,... 1 c = 1 , 1 , 1 , 1 ,... n n 1 2 3 4 1 n 1 1 1 1 dn = , , , ,... − 3 − 3 9 − 27 81 1 1 1 1 1 2 17 43 En =1+ + + + + 1, 2, 2 , 2 , 2 , 2 ,... 1! 2! 3! n! 2 3 24 60 3 2n+1 3 3 5 x n x x x x Sn = T2n+1 sin x = x + +( 1) x,x ,x + ,... { } − 3! − (2n + 1)! − 3! − 3! 5! The last two sequences are derived from the Taylor polynomials of ex (at x = 1) and sin x (at any x). The last example Sn really is a sequence of functions, i.e. for every choice of x you get a diﬀerent sequence.

17.3. Deﬁnition. A sequence of numbers (an)n∞=1 converges to a limit L, if for every ǫ> 0 there is a number Nǫ such that for all n > Nǫ one has

an L < ǫ. | − | One writes lim an = L n →∞ 46

1 17.4. Example: lim = 0. The sequence cn = 1/n converges to 0. To prove this n →∞ n let ǫ> 0 be given. We have to ﬁnd an Nǫ such that

cn < ǫ for all n > Nǫ. | | The cn are all positive, so cn = cn, and hence | | 1 1 cn < ǫ < ǫ n> , | | ⇐⇒ n ⇐⇒ ǫ 1 which prompts us to choose Nǫ = 1/ǫ. The calculation we just did shows that if n> ǫ = Nǫ, then cn < ǫ. That means that limn cn = 0. | | →∞ 17.5. Example: lim an = 0 if a < 1. As in the previous example one can show n | | n →∞ that limn 2− = 0, and more generally, that for any constant a with 1 | | ⇐⇒ ln a | | n The choice Nǫ = (ln ǫ)/(ln a ) therefore guarantees that a < ǫ whenever n > Nǫ. | | | | One can show that the operation of taking limits of sequences obeys the same rules as taking limits of functions.

17.6. Theorem. If

lim an = A and lim bn = B, n n →∞ →∞ then one has

lim an bn = A B n →∞ ± ± lim anbn = AB n →∞ a A lim n = (assuming B = 0). n →∞ bn B The so-called “sandwich theorem” for ordinary limits also applies to limits of sequences. Namely, one has

17.7. “Sandwich theorem”. If an is a sequence which satisﬁes bn < an < cn for all n, and if limn bn = limn cn = 0, then limn an = 0. Finally, one can show →∞ →∞ →∞ this:

17.8. Theorem. If f(x) is a function which is continuous at x = A, and an is a sequence which converges to A, then

lim f(an)= f lim an = f(A). n n →∞ →∞

17.9. Example. Since limn 1/n = 0 and since f(x) = cos x is continuous at →∞ x = 0 we have 1 lim cos = cos 0 = 1. n →∞ n 47

17.10. Example. You can compute the limit of any rational function of n by dividing numerator and denominator by the highest occurring power of n. Here is an example:

2 2n2 1 2 1 2 02 lim − = lim − n = − = 2 n n2 + 3n n 1 + 3 1 1 + 3 02 →∞ →∞ n

1 17.11. Example. [Application of the Sandwich theorem. ] We show that limn = →∞ √n2+1 0 in two diﬀerent ways. Method 1: Since √n2 + 1 > √n2 = n we have 1 1 0 < < . √n2 + 1 n 1 √ 2 The sequences “0” and n both go to zero, so the Sandwich theorem implies that 1/ n + 1 also goes to zero. Method 2: Divide numerator and denominator both by n to get 1/n 1 x an = = f , where f(x)= . 2 n √ 2 1+(1/n) 1+ x 1 Since f(x) is continuous at x = 0, and since 0 as n , we conclude that an n → → ∞ converges to 0.

xn 17.12. Example: lim = 0 for any real number x. If x 1 then this is n n! | | ≤ easy, for we would have x→∞n 1 for all n 0 and thus | |≤ ≥ xn 1 1 1 1 = = n! ≤ n! 1 2 3 (n 1) n ≤ 1 2 2 2 2 2n 1 − − n 1 factors n 1 factors − − n x which shows that limn = 0, by the Sandwich Theorem. →∞ n! For arbitrary x you ﬁrst choose an integer N 2x. Then for all n N one has ≥ ≥ xn x x x x N | | | || | | | use x n! ≤ 1 2 3 n | |≤ 2 n N N N N N 1 ≤ 1 2 3 n 2 Split fraction into two parts, one containing the ﬁrst N factors from both numerator and denominator, the other the remaining factors: N N N N N N N N N N N N N = 1 2 3 N N + 1 n N! N + 1 N + 2 n ≤ N!

=NN /N! <1 <1 <1 Hence we have xn N N 1 n n! ≤ N! 2 if 2 x N and n N. | |≤ ≥ 1 n Here everything is independent of n, except for the last factor ( 2 ) which causes the whole thing to converge to zero as n . →∞ 48

18. Convergence of Taylor Series

18.1. Deﬁnition. Let y = f(x) be some function deﬁned on an interval a

In both cases convergence justiﬁes the idea that you can add inﬁnitely many terms, as suggested by both notations. There is no easy and general criterion which you could apply to a given function f(x) that would tell you if its Taylor series converges for any particular x (except x = 0 – what does the Taylor series look like when you set x = 0?). On the other hand, it turns out that for many functions the Taylor series does converge to f(x) for all x in some interval ρ

18.3. Convergence of the exponential Taylor series. Let f(x) = ex. It turns out the Taylor series of ex converges to ex for every value of x. Here’s why: we had found that 2 n x x x Tne =1+ x + + + , 2! n! and by Lagrange’s formula the remainder is given by xn+1 R ex = eξ , n (n + 1)! where ξ is some number between 0 and x. If x > 0 then 0 <ξ

18.4. The day that all Chemistry stood still. The rate at which a chemical reaction “A B” proceeds depends among other things on the temperature at which the → reaction is taking place. This dependence is described by the Arrhenius law which states that the rate at which a reaction takes place is proportional to

∆E f(T )= e− kT where ∆E is the amount of energy involved in each reaction, k is Boltzmann’s constant, and T is the temperature in degrees Kelvin. If you ignore the constants ∆E and k (i.e. if you set them equal to one by choosing the right units) then the reaction rate is proportional to 1/T f(T )= e− . If you have to deal with reactions at low temperatures you might be inclined to replace this function with its Taylor series at T = 0, or at least the ﬁrst non-zero term in this series. If you were to do this you’d be in for a surprise. To see what happens, let’s look at the following function, 1/x e− x> 0 f(x)= 0 x 0 ≤ This function goes to zero very quickly as x 0. In fact one has → 1/x f(x) e− n t lim = lim = lim t e− = 0. (set t = 1/x) x 0 n x 0 n t ց x ց x →∞ This implies f(x)= o(xn) (x 0) → for any n = 1, 2, 3 . . .. As x 0, this function vanishes faster than any power of x. → 50

The Taylor series

at this point does 1/x y = e− not converge to f

1 2 3

Figure 6. An innocent looking function with an unexpected Taylor series. See example 18.4 which shows that even when a Taylor series of some function f converges you can’t be sure that it converges to f – it could converge to a diﬀerent function.

If you try to compute the Taylor series of f you need its derivatives at x = 0 of all orders. These can be computed (not easily), and the result turns out to be that all derivatives of f vanish at x = 0,

(3) f(0) = f ′(0) = f ′′(0) = f (0) = = 0. The Taylor series of f is therefore x2 x3 T f(x)=0+0 x + 0 + 0 + = 0. ∞ 2! 3! Clearly this series converges (all terms are zero, after all), but instead of converging to the function f(x) we started with, it converges to the function g(x) = 0. What does this mean for the chemical reaction rates and Arrhenius’ law? We wanted to “simplify” the Arrhenius law by computing the Taylor series of f(T ) at T = 0, but we have just seen that all terms in this series are zero. Therefore replacing the Arrhenius reaction rate by its Taylor series at T = 0 has the eﬀect of setting all reaction rates equal to zero.

19. Leibniz’ formulas for ln 2 and π/4

Leibniz showed that 1 1 1 1 1 + + = ln 2 1 − 2 3 − 4 5 − and 1 1 1 1 1 π + + = 1 − 3 5 − 7 9 − 4 Both formulas arise by setting x = 1 in the Taylor series for x2 x3 x4 ln(1 + x)= x + + − 2 3 4 − x3 x5 x7 arctan x = x + + − 3 5 7 − This is only justiﬁed if you show that the series actually converge, which we’ll do here, at least for the ﬁrst of these two formulas. The proof of the second is similar. The following is not Leibniz’ original proof. You begin with the geometric sum 1 ( 1)n+1xn+1 1 x + x2 x3 + +( 1)nxn = + − − − − 1+ x 1+ x 51

Then you integrate both sides from x =0 to x = 1 and get 1 1 1 1 1 dx 1 xn+1dx + +( 1)n = +( 1)n+1 1 − 2 3 − − n + 1 1+ x − 1+ x 0 0 1 xn+1dx =ln2+( 1)n+1 − 1+ x 0 1 k 1 (Use 0 x dx = k+1 .) Instead of computing the last integral you estimate it by saying xn+1 1 xn+1dx 1 1 0 xn+1 = 0 xn+1dx = ≤ 1+ x ≤ ⇒ ≤ 1+ x ≤ n + 2 0 0 Hence 1 xn+1dx lim ( 1)n+1 = 0, n − 1+ x →∞ 0 and we get 1 1 1 1 1 xn+1dx lim + +( 1)n = ln 2 + lim ( 1)n+1 n 1 − 2 3 − − n + 1 n − 1+ x →∞ →∞ 0 = ln 2.

Euler proved that π2 1 1 1 =1+ + + + 6 4 9 n2

20. Proof of Lagrange’s formula

For simplicity assume x> 0. Consider the function (n) f ′′(0) 2 f (0) n n+1 g(t)= f(0) + f ′(0)t + t + + t + Kt f(t), 2 n! − where n f ( )(0) n f(0) + f ′(0)x + + x f(x) (14) K def= n! − − xn+1 We have chosen this particular K to be sure that g(x) = 0. Just by computing the derivatives you also ﬁnd that (n) g(0) = g′(0) = g′′(0) = = g (0) = 0, while (15) g(n+1)(t)=(n + 1)!K f (n+1)(t). − We now apply Rolle’s Theorem n times:

since g(t) vanishes at t = 0 and at t = x there exists an x1 with 0 < x1 < x • such that g′(x1) = 0 since g′(t) vanishes at t = 0 and at t = x1 there exists an x2 with 0

(n+1) We now set ξ = xn+1, and observe that we have shown that g (ξ) = 0, so by (15) we get f (n+1)(ξ) K = . (n + 1)! Apply that to (14) and you ﬁnally get

(n) (n+1) f (0) n f (ξ) n+1 f(x)= f(0) + f ′(0)x + + x + x . n! (n + 1)!

21. Proof of Theorem 16.8

21.1. Lemma. If h(x) is a k times diﬀerentiable function on some interval con- (k 1) taining 0, and if for some integer k < n one has h(0) = h′(0) = = h − (0) = 0, then h(x) h(k)(0) (16) lim = . x 0 k → x k!

Proof. Just apply l’Hopital’s rule k times. You get

0 0 (2) 0 h(x) = h′(x) = h (x) = lim =0 lim =0 lim =0 x 0 xk x 0 kxk 1 x 0 k(k 1)xk 2 → → − → − − (k 1) 0 (k) h − (x) = h (0) = lim =0 x 0 k(k 1) 2x1 k(k 1) 2 1 → − −

First deﬁne the function h(x)= f(x) g(x). If f(x) and g(x) are n times diﬀerentiable, − then so is h(x).

The condition Tnf(x)= Tng(x) means that (n) (n) f(0) = g(0), f ′(0) = g′(0), ..., f (0) = g (0), which says, in terms of h(x),

(n) ( ) h(0) = h′(0) = h′′(0) = = h (0) = 0, † i.e.

Tnh(x) = 0. We now prove the ﬁrst pat of the theorem: suppose f(x) and g(x) have the same nth degree Taylor polynomial. Then we have just argued that Tnh(x) = 0, and Lemma 21.1 n (with k = n) says that limx 0 h(x)/x = 0, as claimed. → n To conclude we show the converse also holds. So suppose that limx 0 h(x)/x = 0. → We’ll show that ( ) follows. If ( ) were not true then there would be a smallest integer † † k n such that ≤ (k 1) (k) h(0) = h′(0) = h′′(0) = = h − (0) = 0, but h (0) = 0. This runs into the following contradiction with Lemma 21.1 (k) n h (0) h(x) h(x) x n k 0 = = lim = lim = 0 lim x − = 0. x 0 k x 0 n k x 0 k! → x → x x → ( ) ∗ Here the limit ( ) exists because n k. ∗ ≥ 53

22. PROBLEMS

TAYLOR’S FORMULA

179. Find a second order polynomial (i.e. 194. Find the nth degree Taylor polyno- a a quadratic function) Q(x) such that mial Tn f(x) of the following functions Q(7) = 43,Q′(7) = 19,Q′′(7) = 11. f(x) 180. Find a second order polynomial p(x) n a f(x) 3 such that p(2) = 3, p′(2) = 8, and 2 0 1+ x x − 3 p′′(2) = 1. 3 0 1+ x x − 3 − 25 0 1+ x x 181. A Third order polynomial P (x) sat- − 3 25 2 1+ x x isfies P (0) = 1, P ′(0) = 3, P ′′(0) = − 3 − 2 1 1+ x x 8, P ′′′(0) = 24. Find P (x). 2− − 1 1 x 182. Let f(x) = √x +25. Find the poly- 2 1 x2 nomial P (x) of degree three such that 5 1 1/x P (k)(0) = f (k)(0) for k = 0, 1, 2, 3. 5 0 1/(1 + x) 3 0 1/(1 3x + 2x2) 183. Let f(x)=1+ x x2 x3. Com- − − − For which of these combinations pute and graph T0f(x), T1f(x), T2f(x), (n, a, f(x)) is T af(x) the same as f(x)? T3f(x), and T4f(x), as well as f(x) itself n (so, for each of these functions find where they are positive or negative, where they ∗ ∗ ∗ are increasing/decreasing, and find the Compute the Taylor series T f(t) for ∞ inflection points on their graph.) the following functions (α is a constant). Give a formula for the coefficient of xn 184. Find T3 sin x and T5 sin x. in T f(t). (Be smart. Remember prop- Graph T sin x and T sin x as well ∞ 3 5 erties of the logarithm, definitions of the as y = sin x in one picture. (As be- hyperbolic functions, partial fraction de- fore, find where these functions are posi- composition.) tive or negative, where they are increas- t ing/decreasing, and find the inflection 195. e points on their graph. This problem 196. eαt can&should be done without a graphing calculator.) 197. sin(3t) a a Compute T0 f(x), T1 f(x) and 198. sinh t T af(x) for the following functions. 2 199. cosh t 185. f(x)= x3, a = 0; then for a = 1 and 1 200. a = 2. 1 + 2t 1 3 186. f(x)= , a = 1. Also do a = 2. 201. x (2 t)2 − 187. f(x)= √x, a = 1. 202. ln(1 + t) 2 188. f(x)=ln x, a = 1. Also a = e . 203. ln(2 + 2t) 189. f(x)=ln √x, a = 1. 204. ln √1+ t 190. f(x) = sin(2x), a = 0, also a = π/4. 205. ln(1 + 2t) 191. f(x) = cos(x), a = π. 1+ t 206. ln 192. f(x)=(x 1)2, a = 0, and also a = 1. 1 t − − 1 1 193. f(x)= , a = 0. 207. [hint:PFD!] ex 1 t2 − 54

t 208. 218. Group problem. The following 1 t2 − questions ask you to rediscover New- 209. sin t + cos t ton’s Binomial Formula, which is just the Taylor series for (1 + x)n. New- 210. 2 sin t cos t ton’s formula generalizes the formulas 211. tan t (3 terms only) for (a + b)2, (a + b)3, etc that you get 2 using Pascal’s triangle. It allows non in- 212. 1+ t2 t4 − 3 teger exponents which are allowed to be 5 either positive and negative. Reread sec- 213. (1 + t) tion 13 before doing this problem. 214. √3 1+ t (a) Find the Taylor series of f(x)= 1/2 x4 (10) √1+ x (= (1+ x) ) 215. f(x)= 2 , what is f (0)? 1+4x (b) Find the coeﬃcient of x4 in the 216. Group problem. Compute the Tay- Taylor series of f(x) = (1+ x)π (don’t lor series of the following two functions do the arithmetic!) f(x) = sin a cos x + cos a sin x (c) Let p be any real number. Com- and pute the terms of degree 0, 1, 2 and 3 of g(x) = sin(a + x) the Taylor series of p where a is a constant. f(x)=(1+ x) 217. Group problem. Compute the Tay- (d) Compute the Taylor polynomial p lor series of the following two functions of degree n of f(x)=(1+ x) . h(x) = cos a cos x sin a sin x (e) Write the result of (d) for the − exponents p = 2, 3 and also, for p = and 1, 2, 3 and ﬁnally for p = 1 . The − − − 2 k(x) = cos(a + x) Binomial Theorem states that this se- where a is a constant. ries converges when x < 1. | |

LAGRANGE’S FORMULA FOR THE REMAINDER

219. Find the fourth degree Taylor polyno- 222. Follow the method of problem 221 to mial T4 cos x for the function f(x) = compute √10: { } cos x and estimate the error cos x | − P4(x) for x < 1. (a) Use Taylor’s formula with | | | f(x) = √9+ x, n = 1, to calculate √10 220. Find the 4th degree Taylor polyno- approximately. Show that the error is mial T4 sin x for the function f(x) = { } less than 1/216. sin x. Estimate the error sin x | − T4 sin x for x < 1. { }| | | (b) Repeat with n = 2. Show that 221. (Computing the cube root of 9) The the error is less than 0.0003. cube root of 8 = 2 2 2 is easy, and 9 × × is only one more than 8. So you could try to compute √3 9 by viewing it as √3 8 + 1. 223. Find the eighth degree Taylor polynomial T8f(x) about the point 0 for the (a) Let f(x) = √3 8+ x. Find 3 function f(x) = cos x and estimate the T2f(x), and estimate the error √9 | − error cos x T8f(x) for x < 1. T2f(1) . | − | | | | (b) Repeat part (i) for “n = 3”, Now ﬁnd the ninth degree Tay- 3 i.e. compute T3f(x) and estimate √9 lor polynomial, and estimate cos x | − | − T3f(1) . T9f(x) for x 1. | | | |≤ 55

LITTLE-OH AND MANIPULATING TAYLOR POLYNOMIALS

Are the following statements True (d) Using a partial fraction de- or False? In mathematics this means composition of g(x) find a formula for (n) that you should either show that the g (0), and hence for gn. statement always holds or else give at 239. Answer the same questions as in the least one counterexample, thereby show- previous problem, for the functions ing that the statement is not always true. x 2 2 h(x)= 224. (1 + x ) 1= o(x)? 2 3 x + x2 − − 225. (1 + x2)2 1= o(x2)? and − 2 x k(x)= − . 226. √1+ x √1 x = o(x) ? 2 3 x + x2 − − − 227. o(x)+ o(x)= o(x)? n 240. Let hn be the coefficient of x in the 228. o(x) o(x)= o(x)? − Taylor series of 229. o(x) o(x)= o(x) ? 1+ x h(x)= 2 . 2 2 2 5x + 2x 230. o(x )+ o(x)= o(x )? − (a) Find a recursion relation for the 231. o(x2) o(x2)= o(x3)? − hn. 232. o(2x)= o(x) ? (b) Compute h0, h1, ..., h8. 2 233. o(x)+ o(x )= o(x)? (c) Derive a formula for hn valid for 2 2 all n, by using a partial fraction expan- 234. o(x)+ o(x )= o(x )? sion. 235. 1 cos x = o(x)? − (d) Is h2009 more or less than a million? A billion? 236. Define Find the Taylor series for the 1/x2 e− x = 0 following functions, by substituting, f(x)= 0 x = 0 adding, multiplying, applying long division and/or differentiating known series This function goes to zero very quickly for 1 , ex, sin x, cos x and ln x. as x 0 but is 0 only at 0. Prove that 1+x → n f(x)= o(x ) for every n. 241. eat 237. For which value(s) of k is √1+ x2 = 242. e1+t k 1+ o(x ) (as x 0)? t2 → 243. e− For which value(s) of k is √3 1+ x2 = 1+ t 1+ o(xk) (as x 0)? 244. → 1 t For which value(s) of k is 1 − − 1 cos x2 = o(xk) (as x 0)? 245. → 1 + 2t 238. Group problem. Let gn be the co- 246. efficient of xn in the Taylor series of the sin(x) if x = 0 function f(x)= x 1 if x = 0 1 g(x)= 2 3 x + x2 ln(1 + x) − 247. (a) Compute g0 and g1 directly from x the definition of the Taylor series. et 248. (b) Show that the recursion relation 1 t − gn = 3gn 1 2gn 2 holds for all n 2. 1 − − − ≥ 249. (c) Compute g2, g3, g4, g5. √1 t − 56

1 t 250. (recommendation: use the 253. T4[e− sin 2t] √1 t2 − 1 answer to problem 249) 254. 2 t t2 251. arcsin t − − 3 2 (use problem 249 again) 255. 1 + 2t + t 2 t 256. ln(1 t ) 252. Compute T4[e− cos t] (See example − 16.11.) 257. sin t cos t

LIMITS OF SEQUENCES

n2 Compute the following limits: 264. lim n (1.01)n n →∞ 258. lim n n 1000 →∞ 2n 3 265. lim − n n! n2 →∞ 259. lim n! + 1 n 266. lim →∞ 2n 3 n − →∞ (n + 1)! n2 267. Group problem. Compute 260. lim 2 n 2n2 + n 3 (n!) →∞ − lim [Hint: write out all the fac- n (2n)! 2n + 1 →∞ 261. lim tors in numerator and denominator.] n n →∞ 1 2 − 268. Group problem. Let fn be the nth 2n + 1 262. lim Fibonacci number. Compute n 1 3n →∞ − fn n lim e + 1 n fn 1 263. lim →∞ − n 1 2n →∞ −

CONVERGENCE OF TAYLOR SERIES

269. Prove that the Taylor series for all x, does it have to converge to f(x), f(x) = cos x converges to f(x) for all or could it converge to some other func- real numbers x (by showing that the re- tion? mainder term goes to zero as n ). →∞ 275. For which real numbers x does the 1 270. Prove that the Taylor series for Taylor series of f(x) = converge g(x) = sin(2x) converges to g(x) for all 1 x − real numbers x . to f(x)? 276. For which real numbers x does the 271. Prove that the Taylor series for 1 h(x) = cosh(x) converges to h(x) for all Taylor series of f(x) = con- 1 x2 real numbers x . verge to f(x)? (hint: a substitution− may 272. Prove that the Taylor series for help.) k(x) = e2x+3 converges to k(x) for all 277. For which real numbers x does the real numbers x . 1 Taylor series of f(x) = converge 1+ x2 273. Prove that the Taylor series for ℓ(x)= to f(x)? cos x π converges to ℓ(x) for all real − 7 numbers x. 278. For which real numbers x does the 1 Taylor series of f(x)= converge 274. Group problem. If the Taylor se- 3 + 2x ries of a function y = f(x) converges for to f(x)? 57

279. For which real numbers x does the 284. Show that the Taylor series for f(x)= Taylor series of f(x)= 1 converge to 1/(1+x3) converges whenever 1

APPROXIMATING INTEGRALS

288. (a) Compute T2 sin t and give an up- Note: You need not ﬁnd p(t) or the in- { } 1 2 per bound for R2 sin t for 0 t 0.5 tegral p(x ) dx. { } ≤ ≤ 0 (b) Use part (a) to approximate 0.5 0.1 sin(x2) dx, and give an upper 290. Approximate arctan xdx and es- 0 0 bound for the error in your approxima- timate the error in your approximation tion. by analyzing T2f(t) and R2f(t) where f(t) = arctan t. 289. (a) Find the second degree Taylor polynomial for the function et. 0.1 2 x2 291. Approximate x e− dx and es- (b) Use it to give an estimate for the 0 integral timate the error in your approximation 1 x2 by analyzing T3f(t) and R3f(t) where e dx t f(t)= te− . 0 (c) Suppose instead we used the 5th 0.5 degree Taylor polynomial p(t) for et to 292. Estimate 1+ x4 dx with an er- 0 give an estimate for the integral: 4 ror of less than 10− . 1 x2 e dx 0.1 0 293. Estimate arctan xdx with an er- Give an upper bound for the error: 0 ror of less than 0.001. 1 1 2 ex dx p(x2) dx − 0 0 58

Chapter 3: Complex Numbers and the Complex Exponential

23. Complex numbers

The equation x2 + 1 = 0 has no solutions, because for any real number x the square x2 is nonnegative, and so x2 + 1 can never be less than 1. In spite of this it turns out to be very useful to assume that there is a number i for which one has (17) i2 = 1. − Any complex number is then an expression of the form a + bi, where a and b are old- fashioned real numbers. The number a is called the real part of a + bi, and b is called its imaginary part. Traditionally the letters z and w are used to stand for complex numbers. Since any complex number is speciﬁed by two real numbers one can visualize them by plotting a point with coordinates (a,b) in the plane for a complex number a + bi. The plane in which one plot these complex numbers is called the Complex plane, or Argand plane.

b = Im(z) z = a + bi

b2 2 + √ a = z = | r | θ = arg z

a = Re(z)

Figure 7. A complex number.

You can add, multiply and divide complex numbers. Here’s how: To add (subtract) z = a + bi and w = c + di z + w =(a + bi)+(c + di)=(a + c)+(b + d)i, z w =(a + bi) (c + di)=(a c)+(b d)i. − − − − 59

To multiply z and w proceed as follows: zw =(a + bi)(c + di) = a(c + di)+ bi(c + di) = ac + adi + bci + bdi2 =(ac bd)+(ad + bc)i − where we have use the deﬁning property i2 = 1 to get rid of i2. − To divide two complex numbers one always uses the following trick. a + bi a + bi c di = − c + di c + di c di − (a + bi)(c di) = − (c + di)(c di) − Now (c + di)(c di)= c2 (di)2 = c2 d2i2 = c2 + d2, − − − so a + bi (ac + bd)+(bc ad)i = − c + di c2 + d2 ac + bd bc ad = + − i c2 + d2 c2 + d2 Obviously you do not want to memorize this formula: instead you remember the trick, i.e. to divide c + di into a + bi you multiply numerator and denominator with c di. − For any complex number w = c+di the number c di is called its complex conjugate. − Notation: w = c + di, w¯ = c di. − A frequently used property of the complex conjugate is the following formula (18) ww¯ =(c + di)(c di)= c2 (di)2 = c2 + d2. − − The following notation is used for the real and imaginary parts of a complex number z. If z = a + bi then a = the Real Part of z = Re(z), b = the Imaginary Part of z = Im(z). Note that both Rez and Imz are real numbers. A common mistake is to say that Imz = bi. The “i” should not be there.

24. Argument and Absolute Value

For any given complex number z = a + bi one defines the absolute value or modulus to be z = a2 + b2, | | so z is the distance from the origin to the point z in the complex plane (see figure 7). | | The angle θ is called the argument of the complex number z. Notation: arg z = θ. The argument is defined in an ambiguous way: it is only defined up to a multiple of 2π. E.g. the argument of 1 could be π, or π, or 3π, or, etc. In general one says − − arg( 1) = π + 2kπ, where k may be any integer. − From trigonometry one sees that for any complex number z = a + bi one has a = z cos θ, and b = z sin θ, | | | | so that z = z cos θ + i z sin θ = z cos θ + i sin θ . | | | | | | | | 60 and sin θ b tan θ = = . cos θ a

24.1. Example: Find argument and absolute value of z = 2+ i. Solution: z = √22 + 12 = √5. z lies in the ﬁrst quadrant so its argument θ is an angle between 0 | | 1 and π/2. From tan θ = 1 we then conclude arg(2 + i)= θ = arctan . 2 2

25. Geometry of Arithmetic

Since we can picture complex numbers as points in the complex plane, we can also try to visualize the arithmetic operations “addition” and “multiplication.” To add z and

z + w

z c b a

Figure 8. Addition of z = a + bi and w = c + di w one forms the parallelogram with the origin, z and w as vertices. The fourth vertex then is z + w. See ﬁgure 8.

iz = b + ai −

z = a + bi

Figure 9. Multiplication of a + bi by i.

To understand multiplication we ﬁrst look at multiplication with i. If z = a + bi then iz = i(a + bi)= ia + bi2 = ai b = b + ai. − − Thus, to form iz from the complex number z one rotates z counterclockwise by 90 degrees. See ﬁgure 9. If a is any real number, then multiplication of w = c + di by a gives aw = ac + adi, 61

3z 2z z

z − 2z −

Figure 10. Multiplication of a real and a complex number so aw points in the same direction, but is a times as far away from the origin. If a < 0 then aw points in the opposite direction. See ﬁgure 10. Next, to multiply z = a + bi and w = c + di we write the product as zw =(a + bi)w = aw + biw. Figure 11 shows a + bi on the right. On the left, the complex number w was ﬁrst drawn,

aw+biw

biw

aw a+bi b iw θ ϕ w θ a

Figure 11. Multiplication of two complex numbers then aw was drawn. Subsequently iw and biw were constructed, and ﬁnally zw = aw+biw was drawn by adding aw and biw. One sees from ﬁgure 11 that since iw is perpendicular to w, the line segment from 0 to biw is perpendicular to the segment from 0 to aw. Therefore the larger shaded triangle on the left is a right triangle. The length of the adjacent side is a w , and the length of | | the opposite side is b w . The ratio of these two lengths is a : b, which is the same as for | | the shaded right triangle on the right, so we conclude that these two triangles are similar. The triangle on the left is w times as large as the triangle on the right. The two | | angles marked θ are equal. Since zw is the length of the hypothenuse of the shaded triangle on the left, it is w | | | | times the hypothenuse of the triangle on the right, i.e. zw = w z . | | | | | | 62

The argument of zw is the angle θ + ϕ; since θ = arg z and ϕ = arg w we get the following two formulas (19) zw = z w | | | | | | (20) arg(zw) = arg z + arg w, in other words, when you multiply complex numbers, their lengths get multiplied and their arguments get added.

26. Applications in Trigonometry

26.1. Unit length complex numbers. For any θ the number z = cos θ + i sin θ has length 1: it lies on the unit circle. Its argument is arg z = θ. Conversely, any complex number on the unit circle is of the form cos φ + i sin φ, where φ is its argument.

26.2. The Addition Formulas for Sine & Cosine. For any two angles θ and φ one can multiply z = cos θ + i sin θ and w = cos φ + i sin φ. The product zw is a complex number of absolute value zw = z w = 1 1, and with argument arg(zw) = | | | | | | arg z + arg w = θ + φ. So zw lies on the unit circle and must be cos(θ + φ)+ i sin(θ + φ). Thus we have (21) (cos θ + i sin θ)(cos φ + i sin φ) = cos(θ + φ)+ i sin(θ + φ). By multiplying out the Left Hand Side we get (22) (cos θ + i sin θ)(cos φ + i sin φ) = cos θ cos φ sin θ sin φ − + i(sin θ cos φ + cos θ sin φ). Compare the Right Hand Sides of (21) and (22), and you get the addition formulas for Sine and Cosine: cos(θ + φ) = cos θ cos φ sin θ sin φ − sin(θ + φ) = sin θ cos φ + cos θ sin φ

26.3. De Moivre’s formula. For any complex number z the argument of its square z2 is arg(z2) = arg(z z) = arg z + arg z = 2 arg z. The argument of its cube is arg z3 = arg(z z2) = arg(z) + arg z2 = arg z + 2 arg z = 3 arg z. Continuing like this one ﬁnds that (23) arg zn = n arg z for any integer n. Applying this to z = cos θ + i sin θ you ﬁnd that zn is a number with absolute value zn = z n = 1n = 1, and argument n arg z = nθ. Hence zn = cos nθ + i sin nθ. So we | | | | have found (24) (cos θ + i sin θ)n = cos nθ + i sin nθ. This is de Moivre’s formula. For instance, for n = 2 this tells us that cos 2θ + i sin 2θ = (cos θ + i sin θ)2 = cos2 θ sin2 θ + 2i cos θ sin θ. − Comparing real and imaginary parts on left and right hand sides this gives you the double angle formulas cos 2θ = cos2 θ sin2 θ and sin 2θ = 2 sin θ cos θ. − For n = 3 you get, using the Binomial Theorem, or Pascal’s triangle, (cos θ + i sin θ)3 = cos3 θ + 3i cos2 θ sin θ + 3i2 cos θ sin2 θ + i3 sin3 θ = cos3 θ 3 cos θ sin2 θ + i(3 cos2 θ sin θ sin3 θ) − − 63 so that cos 3θ = cos3 θ 3 cos θ sin2 θ − and sin 3θ = cos2 θ sin θ sin3 θ. − In this way it is fairly easy to write down similar formulas for sin 4θ, sin 5θ, etc.. . .

27. Calculus of complex valued functions

A complex valued function on some interval I =(a,b) R is a function f : I C. ⊆ → Such a function can be written as in terms of its real and imaginary parts, (25) f(x)= u(x)+ iv(x), in which u, v : I R are two real valued functions. → One deﬁnes limits of complex valued functions in terms of limits of their real and imaginary parts. Thus we say that lim f(x)= L x x → 0 if f(x)= u(x)+ iv(x), L = A + iB, and both lim u(x)= A and lim v(x)= B x x x x → 0 → 0 hold. From this deﬁnition one can prove that the usual limit theorems also apply to complex valued functions.

27.1. Theorem. If limx x f(x)= L and limx x g(x)= M, then one has → 0 → 0 lim f(x) g(x)= L M, x x → 0 ± ± lim f(x)g(x)= LM, x x → 0 f(x) L lim = , provided M = 0. x x → 0 g(x) M

The derivative of a complex valued function f(x)= u(x)+iv(x) is deﬁned by simply diﬀerentiating its real and imaginary parts:

(26) f ′(x)= u′(x)+ iv′(x). Again, one ﬁnds that the sum,product and quotient rules also hold for complex valued functions.

27.2. Theorem. If f,g : I C are complex valued functions which are diﬀerentiable → at some x0 I, then the functions f g, fg and f/g are diﬀerentiable (assuming g(x0) = 0 ∈ ± in the case of the quotient.) One has

(f g)′(x0)= f ′(x0) g′(x0) ± ± (fg)′(x0)= f ′(x0)g(x0)+ f(x0)g′(x0)

f ′ f ′(x0)g(x0) f(x0)g′(x0) (x0)= − g g(x )2 0 Note that the chain rule does not appear in this list! See problem 324 for more about the chain rule.

28. The Complex Exponential Function

We ﬁnally give a deﬁnition of ea+bi. First we consider the case a = 0: 64

eiθ = cos θ + i sin θ

θ 1

Figure 12. Euler’s deﬁnition of eiθ

28.1. Deﬁnition. For any real number t we set eit = cos t + i sin t. See Figure 12.

28.2. Example. eπi = cos π + i sin π = 1. This leads to Euler’s famous formula − eπi +1=0, which combines the ﬁve most basic quantities in mathematics: e, π, i, 1, and 0.

Reasons why the definition 28.1 seems a good definition. Reason 1. We haven’t defined eit before and we can do anything we like. Reason 2. Substitute it in the Taylor series for ex: (it)2 (it)3 (it)4 eit =1+ it + + + + 2! 3! 4! t2 t3 t4 t5 =1+ it i + + i − 2! − 3! 4! 5! − = 1 t2/2! + t4/4! − − + i t t3/3! + t5/5! − − = cos t + i sin t. This is not a proof, because before we had only proved the convergence of the Taylor series for ex if x was a real number, and here we have pretended that the series is also good if you substitute x = it. Reason 3. As a function of t the definition 28.1 gives us the correct derivative. Namely, using the chain rule (i.e. pretending it still applies for complex functions) we would get deit = ieit. dt Indeed, this is correct. To see this proceed from our definition 28.1: deit d cos t + i sin t = dt dt d cos t d sin t = + i dt dt = sin t + i cos t − = i(cos t + i sin t) 65

Reason 4. The formula ex ey = ex+y still holds. Rather, we have eit+is = eiteis. To check this replace the exponentials by their deﬁnition:

eiteis = (cos t + i sin t)(cos s + i sin s) = cos(t + s)+ i sin(t + s)= ei(t+s).

Requiring ex ey = ex+y to be true for all complex numbers helps us decide what ea+bi shoud be for arbitrary complex numbers a + bi.

28.3. Deﬁnition. For any complex number a + bi we set

ea+bi = ea eib = ea(cos b + i sin b). One veriﬁes as above in “reason 3” that this gives us the right behaviour under diﬀeren- tiation. Thus, for any complex number r = a + bi the function

y(t)= ert = eat(cos bt + i sin bt) satisﬁes rt de rt y′(t)= = re . dt

29. Complex solutions of polynomial equations

29.1. Quadratic equations. The well-known quadratic formula tells you that the equation

(27) ax2 + bx + c = 0 has two solutions, given by

b √D (28) x = − ± , D = b2 4ac. ± 2a − If the coeﬃcients a,b,c are real numbers and if the discriminant D is positive, then this formula does indeed give two real solutions x+ and x . However, if D< 0, then there are − no real solutions, but there are two complex solutions, namely b √ D x = − i − ± 2a ± 2a

29.2. Example: solve x2 + 2x +5=0. Solution: Use the quadratic formula, or complete the square:

x2 + 2x +5=0 x2 + 2x +1= 4 ⇐⇒ − (x + 1)2 = 4 ⇐⇒ − x +1= 2i ⇐⇒ ± x = 1 2i. ⇐⇒ − ± So, if you allow complex solutions then every quadratic equation has two solutions, unless the two solutions coincide (the case D = 0, in which there is only one solution.) 66

1 + i √3 1 + i √3 − 2 2 2 2

1 1 −

1 i √3 1 i √3 − 2 − 2 2 − 2

Figure 13. The sixth roots of 1. There are six of them, and they re arranged in a regular hexagon.

29.3. Complex roots of a number. For any given complex number w there is a method of ﬁnding all complex solutions of the equation

(29) zn = w if n = 2, 3, 4, is a given integer. To ﬁnd these solutions you write w in polar form, i.e. you ﬁnd r > 0 and θ such that w = reiθ. Then z = r1/neiθ/n is a solution to (29). But it isn’t the only solution, because the angle θ for which w = reiθ isn’t unique – it is only determined up to a multiple of 2π. Thus if we have found one angle θ for which w = riθ , then we can also write

w = rei(θ+2kπ), k = 0, 1, 2, ± ± The nth roots of w are then θ k 1/n i n +2 n π zk = r e Here k can be any integer, so it looks as if there are inﬁnitely many solutions. However, if you increase k by n, then the exponent above increases by 2πi, and hence zk does not change. In a formula:

zn = z0, zn+1 = z1, zn+2 = z2, ... zk+n = zk

So if you take k = 0, 1, 2, , n 1 then you have had all the solutions. − The solutions zk always form a regular polygon with n sides. 67

29.4. Example: ﬁnd all sixth roots of w = 1. We are to solve z6 = 1. First write 1 in polar form,

1 = 1 e0i = 1 e2kπi, (k = 0, 1, 2,...). ± ± Then we take the 6th root and ﬁnd

1/6 2kπi/6 kπi/3 zk = 1 e = e , (k = 0, 1, 2,...). ± ± The six roots are

πi/3 1 i 2πi/3 1 i z0 = 1 z1 = e = + √3 z2 = e = + √3 2 2 − 2 2 4πi/3 1 i 5πi/3 1 i z3 = 1 z4 = e = √3 z5 = e = √3 − − 2 − 2 2 − 2

30. Other handy things you can do with complex numbers

30.1. Partial fractions. Consider the partial fraction decomposition

x2 + 3x 4 A Bx + C − = + (x 2)(x2 + 4) x 2 x2 + 4 − − The coefficient A is easy to find: multiply with x 2 and set x = 2 (or rather, take the − limit x 2) to get → 22 + 3 2 4 A = − = . 22 + 4 Before we had no similar way of finding B and C quickly, but now we can apply the same trick: multiply with x2 + 4,

x2 + 3x 4 A − = Bx + C +(x2 + 4) , (x 2) x 2 − − and substitute x = 2i. This make x2 + 4 = 0, with result

(2i)2 + 3 2i 4 − = 2iB + C. (2i 2) − Simplify the complex number on the left:

(2i)2 + 3 2i 4 4 + 6i 4 − = − − (2i 2) 2 + 2i − − 8 + 6i = − 2 + 2i − ( 8 + 6i)( 2 2i) = − − − ( 2)2 + 22 − 28 + 4i = 8 7 i = + 2 2

7 i So we get 2iB + C = 2 + 2 ; since B and C are real numbers this implies 1 7 B = , C = . 4 2 68

30.2. Certain trigonometric and exponential integrals. You can compute

I = e3x cos 2xdx by integrating by parts twice. You can also use that cos 2x is the real part of e2ix. Instead of computing the real integral I, we look at the following related complex integral

J = e3xe2ixdx which we get from I by replacing cos 2x with e2ix. Since e2ix = cos 2x + i sin 2x we have

J = e3x(cos 2x + i sin 2x)dx = e3x cos 2xdx + i e3x sin 2xdx i.e., J = I + something imaginary. The point of all this is that J is easier to compute than I: e(3+2i)x J = e3xe2ixdx = e3x+2ixdx = e(3+2i)xdx = + C 3 + 2i where we have used that 1 eaxdx = eax + C a holds even if a is complex is a complex number such as a =3+2i. To ﬁnd I you have to compute the real part of J, which you do as follows: e(3+2i)x cos 2x + i sin 2x = e3x 3 + 2i 3 + 2i (cos 2x + i sin 2x)(3 2i) = e3x − (3+2i)(3 2i) − 3 cos 2x + 2 sin 2x + i( ) = e3x 13 so 3x 3x 3 2 e cos 2xdx = e 13 cos 2x + 13 sin 2x + C. 30.3. Complex amplitudes. A harmonic oscillation is given by y(t)= A cos(ωt φ), − where A is the amplitude, ω is the frequency, and φ is the phase of the oscillation. If you add two harmonic oscillations with the same frequency ω, then you get another harmonic oscillation with frequency ω. You can prove this using the addition formulas for cosines, but there’s another way using complex exponentials. It goes like this. Let y(t) = A cos(ωt φ) and z(t) = B cos(ωt θ) be the two harmonic oscillations − − we wish to add. They are the real parts of iωt iφ iφ iωt Y (t)= A cos(ωt φ)+ i sin(ωt φ) = Ae − = Ae− e { − − } iωt iθ iθ iωt Z(t)= B cos(ωt θ)+ i sin(ωt θ) = Be − = Be− e { − − } Therefore y(t)+ z(t) is the real part of Y (t)+ Z(t), i.e. y(t)+ z(t)= Re Y (t) + Re Z(t) = Re Y (t)+ Z(t) . The quantity Y (t)+ Z(t) is easy to compute: iφ iωt iθ iωt iφ iθ iωt Y (t)+ Z(t)= Ae− e + Be− e = Ae− + Be− e . If you now do the complex addition iφ iθ iψ Ae− + Be− = Ce− , 69

i.e. you add the numbers on the right, and compute the absolute value C and argument i(ωt ψ) ψ of the sum, then we see that Y (t)+ Z(t)= Ce − . Since we were looking for the − real part of Y (t)+ Z(t), we get y(t)+ z(t)= A cos(ωt φ)+ B cos(ωt θ)= C cos(ωt ψ). − − − iφ iθ iψ The complex numbers Ae− , Be− and Ce− are called the complex amplitudes for the harmonic oscillations y(t), z(t) and y(t)+ z(t). The recipe for adding harmonic oscillations can therefore be summarized as follows: Add the complex amplitudes.

31. PROBLEMS

COMPUTING AND DRAWING COMPLEX NUMBERS

294. Compute the following complex num- (1 + i)(1+2i)(1+3i); bers by hand. 1 √ i √ 2 1 i √ 3 ( 2 2+ 2 2) ; ( 2 + 2 3) ; Draw all numbers in the complex 1 ; 5/(2 i); (or “Argand”) plane (use graph paper or 1+ i − quad paper if necessary). 295. Simplify your answer. Compute absolute value and argument of all numbers involved. For z =2+3i ﬁnd: • 2 2 3 4 (1) z i ; i ; i ; 1/i; (2) z (1+2i)(2 i); (3) z − | | 70

1 (4) z 300. Verify directly from the definition of For z = 2e3i find: addition and multiplication of complex • (1) arg(z) numbers that (2) z | | (a) z + w = w + z (3) z2 (4) 1 (b) zw = wz z π i For z = πe 2 find: (c) z(v + w)= zv + zw • − (1) z | | holds for all complex numbers v,w, and (2) arg(z) z. 296. Plot the following four points in the 301. True or False? (In mathematics this complex plane. Be sure and label them. means that you should either give a

π proof that the statement is always true, 5 i P = √2 e 4 Q =1+2i or else give a counterexample, thereby 1 R = 1 + 2i Z = 1+2i showing that the statement is not always 297. [Deriving the addition formula true.) for tan(θ + φ)] Group problem. Let For any complex numbers z and w θ,φ π , π be two angles. one has ∈ − 2 2 (a) What are the arguments of (a) Re(z)+ Re(w)= Re(z + w) z =1+ i tan θ and w =1+ i tan φ? (b) z + w =z ¯ +w ¯ (Draw both z and w.) (c) Im(z)+ Im(w)= Im(z + w) (b) Compute zw. (d) zw = (¯z)(w ¯) (c) What is the argument of zw? (e) Re(z)Re(w)= Re(zw) (d) Compute tan(arg zw). (f) z/w = (¯z)/(w ¯) 298. Find formulas for cos 4θ, sin 4θ, cos 5θ (g) Re(iz)= Im(z) and sin 6θ in terms of cos θ and sin θ, by using de Moivre’s formula. (h) Re(iz)= iRe(z) 3 (i) Re(iz)= iIm(z) 299. In the following picture draw 2w, 4 w, iw, 2iw, (2+ i)w and (2 i)w. (Try to (j) Im(iz)= Re(z) − − make a nice drawing, use a ruler.) (k) Re(¯z)= Re(z) Make a new copy of the picture, and draww ¯, w¯ and w. 302. Group problem. The imaginary − − part of a complex number is known to be Make yet another copy of the draw- twice its real part. The absolute value ing. Draw 1/w, 1/w¯, and 1/w. For − of this number is 4. Which number is this drawing you need to know where the this? unit circle is in your drawing: Draw a circle centered at the origin with radius 303. The real part of a complex number is of your choice, and let this be the unit known to be half the absolute value of circle. [Depending on which circle you that number. The imaginary part of the draw you will get a diﬀerent answer!] number is 1. Which number is it?

THE COMPLEX EXPONENTIAL

πi 2 πi/2 1 e− e − 304. Compute and draw the following ; ; numbers in the complex plane eπi/4 eπi/4 eπi/4

eπi/3; eπi/2; √2e3πi/4; e17πi/4. e2009πi; e2009πi/2 .

πi i ln 2 4πi/3 πi πi e + 1; e . 8e ; 12e + 3e− . − 71

iβt iβt 305. Compute the absolute value and ar- If Ae + Be− = 2 cos βt + gument of e(ln 2)(1+i). 3 sin βt, then what are A and B? 306. Group problem. Suppose z can be 311. Group problem. (a) Show that you any complex number. can write a “cosine-wave” with ampli- (a) Is it true that ez is always a pos- tude A and phase φ as follows itive number? A cos(t φ)= Re zeit , (b) Is it true that ez = 0? − where the “complex amplitude” is given iφ 307. Group problem. Verify directly by z = Ae− . (See 30.3). from the deﬁnition that § (b) Show that a “sine-wave” with it 1 e− = amplitude A and phase φ as follows eit holds for all real values of t. A sin(t φ)= Re zeit , − 308. Show that where the “complex amplitude” is given it it it it iφ e + e− e e− by z = iAe− . cos t = , sin t = − − 2 2i 312. Find A and φ where A cos(t φ) = − 309. Show that 2 cos(t)+2cos(t 2 π). − 3 1 cosh x = cos ix, sinh x = sin ix. 313. Find A and φ where A cos(t φ) = − i 12 cos(t 1 π) + 12 sin(t 1 π). − 6 − 3 310. The general solution of a second or- 314. Find A and φ where A cos(t φ) = der linear diﬀerential equation contains − iβt iβt 12 cos(t π/6) + 12 cos(t π/3). expressions of the form Ae + Be− . − − These can be rewritten as C1 cos βt + 315. Find A and φ such that A cos(t φ)= 1 2 − C2 sin βt. cos t π + √3 cos t π . − 6 − 3 REAL AND COMPLEX SOLUTIONS OF ALGEBRAIC EQUATIONS

316. Find and draw all real and complex (e) z4 + 2z2 3 = 0 − solutions of (f) 3z6 = z3 + 2 (a) z2 + 6z + 10 = 0 (g) z5 32 = 0 (b) z3 +8=0 − 5 (c) z3 125 = 0 (h) z 16z = 0 − − (d) 2z2 + 4z +4=0 (i) z4 + z2 12 = 0 −

CALCULUS OF COMPLEX VALUED FUNCTIONS

317. Compute the derivatives of the fol- 318. (a) Compute lowing functions cos 2x 4 dx 1 f(x)= x + i 1 iθ iθ by using cos θ = 2 (e + e− ) and ex- g(x) = log x + i arctan x panding the fourth power. ix2 h(x)= e (b) Assuming a R, compute ∈

2x 2 Try to simplify your answers. e− sin ax dx. 72

π (same trick: write sin ax in terms of com- (f) sin(mx) sin(nx)dx plex exponentials; make sure your final 0 answer has no complex numbers.) These integrals are basic to the theory of Fourier series, which occurs iα iα 319. Use cos α = (e + e− )/2, etc. to in many applications, especially in the evaluate these indefinite integrals: study of wave motion (light, sound, eco- 2 nomic cycles, clocks, oceans, etc.). They (a) cos x dx say that different frequency waves are “independent”. (b) cos4 x dx, 321. Show that cos x+sin x = C cos(x+β) for suitable constants C and β and use (c) cos2 x sin x dx, this to evaluate the following integrals. dx 3 (a) (d) sin x dx, cos x + sin x dx 2 2 (b) (e) cos x sin x dx, (cos x + sin x)2 dx 6 (c) (f) sin x dx A cos x + B sin x where A and B are any constants. (g) sin(3x) cos(5x)dx 322. Group problem. Compute the integrals 2 (h) sin (2x) cos(3x)dx π/2 sin2 kx sin2 lx dx, π/4 0 (i) sin(3x) cos(x)dx where k and l are positive integers. 0 π/3 3 2 323. Group problem. Show that for any (j) sin (x) cos (x)dx integers k,l,m 0 π π/2 (k) sin2(x) cos2(x)dx sin kx sin lx sin mx dx = 0 0 0 π/3 if and only if k + l + m is even. 2 (l) sin(x) cos (x)dx 324. Group problem. (i) Prove the fol- 0 lowing version of the Chain rule: If 320. Group problem. Compute the fol- f : I C is a differentiable complex → lowing integrals when m = n are distinct valued function, and g : J I is a → integers. differentiable real valued function, then 2π h = f g : J C is a differentiable (a) sin(mx) cos(nx)dx ◦ → function, and one has 0 2π h′(x)= f ′(g(x))g′(x). (b) sin(nx) cos(nx)dx 0 (ii) Let n 0 be a nonnegative integer. 2π ≥ Prove that if f : I C is a differentiable (c) cos(mx) cos(nx)dx → function, then g(x) = f(x)n is also dif- 0 π ferentiable, and one has (d) cos(mx) cos(nx)dx n 1 g′(x)= nf(x) − f ′(x). 0 2π Note that the chain rule from part (a) (e) sin(mx) sin(nx)dx does not apply! Why? 0 Complex roots of real polynomials

325. For a and b complex numbers show that 73

(a) a + b = a + b (b) a b = a b (c) a is real iﬀ a = a 2 n 326. For p(x) = a0 + a1x + a2x + anx a polynomial and z a complex number, show that 2 n p(z)= a0 + a1 z + a2 (z) + an (z) 327. For p a real polynomial, i.e., the coeﬃcients ak of p are real numbers, if z is a complex root of p, i.e., p(z) = 0, show z is also a root of p. Hence the complex roots of p occur in conjugate pairs. 328. Using the quadratic formula show directly that the roots of a real quadratic are either both real or a complex conjugate pair. 329. Show that 2+3i and its conjugate 2 3i are the roots of a real polynomial. − 330. Show that for every complex number a there is a real quadratic whose roots are a and a. 331. The Fundamental theorem of Algebra states that every complex polynomial of degree n can be completely factored as a constant multiple of

(x al1)(x α2) (x αn) − − − (The αi may not be distinct.) It was proved by Gauss. Proofs of it are given in courses on Complex Analysis. Use the Fundamental Theorem of Algebra to show that every real polynomial can be factored into a product real polynomials, each of degree 1 or 2. 74

Chapter 4: Diﬀerential Equations

32. What is a DiﬀEq?

A differential equation is an equation involving an unknown function and its derivatives. The order of the differential equation is the order of the highest derivative which appears. A linear differential equation is one of form (n) (n 1) y + a1(x)y − + + an 1(x)y′ + an(x)y = k(x) − where the coefficients a1(x),...,an(x) and the right hand side k(x) are given functions of x and y is the unknown function. Here dky y(k) = dxk denotes the kth derivative of y so this equation has order n. We shall mainly study the case n = 1 where the equation has form

y′ + a(x)y = k(x) and the case n = 2 with constant coeﬃcients where the equation has form

y′′ + ay′ + by = k(x). When the right hand side k(x) is zero the equation is called homogeneous linear and otherwise it is called inhomogeneous linear (or nonhomogeneous linear by some people). For a homogeneous linear equation the sum of two solutions is a solution and a constant multiple of a solution is a solution. This property of linear equations is called the principle of superposition.

33. First Order Separable Equations

A separable differential equation is a diffeq of the form dy (30) y′(x)= F (x)G(y(x)), or = F (x)G(y). dx To solve this equation divide by G(y(x)) to get 1 dy (31) = F (x). G(y(x)) dx Next find a function H(y) whose derivative with respect to y is 1 dy (32) H′(y)= solution: H(y)= . G(y) G(y) Then the chain rule implies that (31) can be written as dH(y(x)) = F (x). dx In words: H(y(x)) is an antiderivative of F (x), which means we can find H(y(x)) by integrating F (x):

(33) H(y(x)) = F (x)dx + C. 75

Once you’ve found the integral of F (x) this gives you y(x) in implicit form: the equation (33) gives you y(x) as an implicit function of x. To get y(x) itself you must solve the equation (33) for y(x). A quick way of organizing the calculation goes like this: dy To solve = F (x)G(y) you ﬁrst separate the variables, dx dy = F (x)dx, G(y) and then integrate, dy = F (x)dx. G(y) The result is an implicit equation for the solution y with one undetermined integration constant. Determining the constant. The solution you get from the above procedure contains an arbitrary constant C. If the value of the solution is speciﬁed at some given x0, i.e. if y(x0) is known then you can express C in terms of y(x0) by using (33). A snag: You have to divide by G(y) which is problematic when G(y) = 0. This has as consequence that in addition to the solutions you found with the above procedure, there are at least a few more solutions: the zeroes of G(y) (see Example 33.2 below). In addition to the zeroes of G(y) there sometimes can be more solutions, as we will see in Example 35.2 on “Leaky Bucket Dating.”

33.1. Example. We solve dz = (1+ z2) cos t. dt Separate variables and integrate dz = cos t dt, 1+ z2 to get arctan z = sin t + C. Finally solve for z and you ﬁnd the general solution z(t) = tan sin(t)+ C . 33.2. Example: The snag in action. If you apply the method to y′(x) = Ky with K a constant, you get y(x)= eK(x+C). No matter how you choose C you never get the function y(x) = 0, even though y(x) = 0 satisﬁes the equation. This is because here G(y)= Ky, and G(y) vanishes for y = 0.

34. First Order Linear Equations

There are two systematic methods which solve a ﬁrst order linear inhomogeneous equation dy + a(x)y = k(x). ( ) dx ‡ You can multiply the equation with an “integrating factor”, or you do a substitution y(x)= c(x)y0(x), where y0 is a solution of the homogeneous equation (that’s the equation you get by setting k(x) 0). ≡ 76

34.1. The Integrating Factor. Let

A(x)= a(x)dx, m(x)= eA(x). Multiply the equation ( ) by the “integrating factor” m(x) to get ‡ dy m(x) + a(x)m(x)y = m(x)k(x). dx By the chain rule the integrating factor satisﬁes dm(x) = A′(x)m(x)= a(x)m(x). dx Therefore one has dm(x)y dy dy = m(x) + a(x)m(x)y = m(x) + a(x)y = m(x)k(x). dx dx dx Integrating and then dividing by the integrating factor gives the solution 1 y = m(x)k(x)dx + C . m(x) In this derivation we have to divide by m(x), but since m(x)= eA(x) and since exponentials never vanish we know that m(x) = 0, no matter which problem we’re doing, so it’s OK, we can always divide by m(x).

34.2. Variation of constants for 1st order equations. Here is the second method of solving the inhomogeneous equation ( ). Recall again that the homogeneous equation ‡ associated with ( ) is ‡ dy + a(x)y = 0. ( ) dx † The general solution of this equation is

A(x) y(x)= Ce− . where the coeﬃcient C is an arbitrary constant. To solve the inhomogeneous equation ( ) ‡ we replace the constant C by an unknown function C(x), i.e. we look for a solution in the form def A(x) y = C(x)y0(x) where y0(x) = e− . (This is how the method gets its name: we are allowing the constant C to vary.)

Then y′ (x)+ a(x)y0(x) = 0 (because y0(x) solves ( )) and 0 † y′(x)+ a(x)y(x)= C′(x)y0(x)+ C(x)y0′ (x)+ a(x)C(x)y0(x)= C′(x)y0(x) so y(x)= C(x)y0(x) is a solution if C′(x)y0(x)= k(x), i.e. k(x) C(x)= dx. y (x) 0 1 Once you notice that y (x)= , you realize that the resulting solution 0 m(x) k(x) y(x)= C(x)y (x)= y (x) dx 0 0 y (x) 0 is the same solution we found before, using the integrating factor. Either method implies the following: 77

34.3. Theorem. The initial value problem dy + a(x)y = 0, y(0) = y , dx 0 has exactly one solution. It is given by x A(x) y = y0e− , where A(x)= a(t)dt. 0 The theorem says three things: (1) there is a solution, (2) there is a formula for the solution, (3) there aren’t any other solutions (if you insist on the initial value y(0) = y0.) The last assertion is just as important as the other two, so I’ll spend a whole section trying to explain why.

35. Dynamical Systems and Determinism

A differential equation which describes how something (e.g. the position of a particle) evolves in time is called a dynamical system. In this situation the independent variable is time, so it is customary to call it t rather than x; the dependent variable, which depends on time is often denoted by x. In other words, one has a differential equation for a function x = x(t). The simplest examples have form dx (34) = f(x,t). dt In applications such a differential equation expresses a law according to which the quantity x(t) evolves with time (synonyms: “evolutionary law”, “dynamical law”, “evolution equation for x”). A good law is deterministic, which means that any solution of (34) is completely determined by its value at one particular time t0: if you know x at time t = t0, then the “evolution law” (34) should predict the values of x(t) at all other times, both in the past (tt0). Our experience with solving differential equations so far ( 33 and 34) tells us that § § the general solution to a differential equation like (34) contains an unknown integration constant C. Let’s call the general solution x(t; C) to emphasize the presence of this constant. If the value of x at some time t0 is known to be, say, x0, then you get an equation

(35) x(t0; C)= x0 which you can try to solve for C. If this equation always has exactly one solution C then the evolutionary law (34) is deterministic (the value of x(t0) always determines x(t) at all other times t); if for some prescribed value x0 at some time t0 the equation (35) has several solutions, then the evolutionary law (34) is not deterministic (because knowing x(t) at time t0 still does not determine the whole solution x(t) at times other than t0).

35.1. Example: Carbon Dating. Suppose we have a fossil, and we want to know how old it is.

All living things contain carbon, which naturally occurs in two isotopes, C14 (unsta- ble) and C12 (stable). A long as the living thing is alive it eats & breaths, and its ratio of C12 to C14 is kept constant. Once the thing dies the isotope C14 decays into C12 at a steady rate.

Let x(t) be the ratio of C14 to C12 at time t. The laws of radioactive decay says that there is a constant k > 0 such that dx(t) = kx(t). dt − 78

Solve this differential equation (it is both separable and first order linear: you choose your method) to find the general solution

kt x(t; C)= Ce− .

After some lab work it is found that the current C14/C12 ratio of our fossil is xnow. Thus we have ktnow tnow xnow = Ce− = C = xnowe . ⇒ Therefore our fossil’s C14/C12 ratio at any other time t is/was

k(tnow t) x(t)= xnowe − .

This allows you to compute the time at which the fossil died. At this time the C14/C12 ratio must have been the common value in all living things, which can be measured, let’s call it xlife. So at the time tdemise when our fossil became a fossil you would have had x(tdemise)= xlife. Hence the age of the fossil would be given by

k(tnow tdemise) 1 xlife xlife = x(tdemise)= xnowe − = tnow tdemise = ln ⇒ − k xnow

35.2. Example: On Dating a Leaky Bucket. A bucket is ﬁlled with water. There’s a hole in the bottom of the bucket so the water streams out at a certain rate.

area = A h(t) the height of water in the bucket h(t) A area of cross section of bucket v a area of hole in the bucket v velocity with which water goes through the hole.

The amount of water in the bucket is A h(t); × The rate at which water is leaving the bucket is a v(t); × Hence dAh(t) = av(t). dt − In fluid mechanics it is shown that the velocity of the water as it passes through the hole only depends on the height h(t) of the water, and that, for some constant K, v(t)= Kh(t). The last two equations together give a differential equation for h(t), namely, dh(t) a = Kh(t). dt − A a √ To make things a bit easier we assume that the constants are such that A K = 2. Then h(t) satisfies

(36) h′(t)= 2 h(t). − This equation is separable, and when you solve it you get dh = 1 = h(t)= t + C. 2√h − ⇒ − 79

h(t)

t -3 -2 -1 1 2 3

Figure 14. Several solutions h(t; C) of the Leaking Bucket Equation (36). Note how they all have the same values when t 1. ≥

This formula can’t be valid for all values of t, for if you take t>C, the RHS becomes negative and can’t be equal to the square root in the LHS. But when t C we do get a ≤ solution, h(t; C)=(C t)2. − This solution describes a bucket which is losing water until at time C it is empty. Motivated by the physical interpretation of our solution it is natural to assume that the bucket stays empty when t>C, so that the solution with integration constant C is given by (C t)2 when t C h(t)= − ≤ 0 for t>C.

We now come to the question: is the Leaky Bucket Equation deterministic? The answer is: NO. If you let C be any negative number, then h(t; C) describes the water level of a bucket which long ago had water, but emptied out at time C < 0. In particular, for all these solutions of the diﬀeq (36) you have h(0) = 0, and knowing the value of h(t) at t = 0 in this case therefore doesn’t tell you what h(t) is at other times. Once you put it in terms of the physical interpretation it is actually quite obvious why this system can’t be deterministic: it’s because you can’t answer the question “If you know that the bucket once had water and that it is empty now, then how much water did it hold one hour ago?”

36. Higher order equations

After looking at ﬁrst order diﬀerential equations we now turn to higher order equations.

36.1. Example: Spring with a weight. A body of mass m is suspended by a spring. There are two forces on the body: gravity and the tension in the spring. Let F y(t) be the sum of these two forces. Newton’s law says that the motion of the weight satisﬁes F = ma where a is the acceleration. The force of gravity is mg where g=32ft/sec2; the quantity mg is called the weight of the body. We assume Hooke’s law which says that Fspring the tension in the spring is proportional to the amount by which the spring is stretched; m the constant or proportionality is called the spring constant. We write k for this spring F constant. gravity 80

The total force acting on the body is therefore F = mg ky(t). − According to Newton’s first/second/third law the acceleration a of the body satisfies F = ma. Since the acceleration a is the second derivative of position y we get the following differential equation for y(t) d2y (37) m = mg ky(t). dt2 − 36.2. Example: the pendulum.

θ L

Fstring

F =mg gravity θ

dθ 2 2 The velocity of the weight on the pendulum is L dt , hence its acceleration is a = Ld θ/dt . There are two forces acting on the weight: gravity (strength mg; direction vertically down) and the tension in the string (strength: whatever it takes to keep the weight on the circle of radius L and center P ; direction parallel to the string). Together they leave a force of size Fgravity sin θ which accelerates the weight. By Newton’s “F = ma” law you get d2θ mL = mg sin θ(t), dt2 − or, canceling ms, d2θ g (38) + sin θ(t) = 0. dt2 L

37. Constant Coeﬃcient Linear Homogeneous Equations

37.1. Differential operators. In this section we study the homogeneous linear differential equation (n) (n 1) (39) y + a1y − + + an 1y′ + any = 0 − where the coefficients a1,...,an are constants. 81

37.2. Examples. The three equations dy y = 0, dx − y′′ y = 0, y′′ + y = 0 − y(iv) y = 0 − are homogeneous linear diﬀerential equations with constant coeﬃcients. Their degrees are 1, 2, 2, and 4. It will be handy to have an abbreviation for the Left Hand Side in (39), so we agree to write L[y] for the result of substituting a function y in the LHS of (39). In other words, for any given function y = y(x) we set

def (n) (n 1) L[y](x) = y (x)+ a1y − (x)+ + an 1y′(x)+ any(x). − We call L an operator. An operator is like a function in that you give it an input, it does a computation and gives you an output. The diﬀerence is that ordinary functions take a number as their input, while the operator L takes a function y(x) as its input, and gives another function (the LHS of (39)) as its output. Since the computation of L[y] involves taking derivatives of y, the operator L is called a diﬀerential operator.

37.3. Example. The diﬀerential equations in the previous example correspond to the diﬀerential operators

L1[y]= y′ y, − L2[y]= y′′ y, L3[y]= y′′ + y − (iv) L4[y]= y y. − So one has d2 sin 2x L3[sin 2x]= + sin 2x = 4 sin 2x + sin 2x = 3 sin 2x. dx2 − −

37.4. The superposition principle. The following theorem is the most important property of linear diﬀerential equations.

37.5. Superposition Principle. For any two functions y1 and y2 we have

L[y1 + y2]= L[y1]+ L[y2]. For any function y and any constant c we have L[cy]= cL[y]. The proof, which is rather straightforward once you know what to do, will be given in lecture. It follows from this theorem that if y1,..., yk are given functions, and c1,..., ck are constants, then

L[c1y1 + + ckyk]= c1L[y1]+ + ckL[yk]. The importance of the superposition principle is that it allows you to take old solutions to the homogeneous equation and make new ones. Namely, if y1, ..., yk are solutions to the homogeneous equation L[y] = 0, then so is c1y1 + +ckyk for any choice of constants c1, ..., ck. 82

37.6. Example. Consider the equation

y′′ 4y = 0. − 2x 2x My cousin Bruce says that the two functions y1(x) = e and y2(x) = e− both are solutions to this equations. You can check that Bruce is right just by substituting his solutions in the equation. The Superposition Principle now implies that 2x 2x y(x)= c1e + c2e− also is a solution, for any choice of constants c1, c2.

37.7. The characteristic polynomial. This example contains in it the general method for solving linear constant coeﬃcient ODEs. Suppose we want to solve the equation (39), i.e. def (n) (n 1) L[y] = y + a1y − + + an 1y′ + any = 0. − Then the ﬁrst thing to do is to see if there are any exponential functions y = erx which satisfy the equation. Since derx d2erx d3erx = rerx, = r2erx, = r3erx, etc. . . . dx dx2 dx3 we see that rx n n 1 rx (40) L[e ]= r + a1r − + an 1r + an e . − The polynomial n n 1 P (r)= r + a1r − + + an 1r + an. − is called the characteristic polynomial. We see that y = erx is a solution of L[y] = 0 if and only if P (r) = 0.

37.8. Example. We look for all exponential solutions of the equation

y′′ 4y = 0. − Substitution of y = erx gives 2 rx rx 2 rx y′′ 4y = r e 4e = r 4 e . − − − rx 2 The exponential e can’t vanish, so y′′ 4y = 0 will hold exactly when r 4 = 0, − − i.e. when r = 2. Therefore the only exponential functions which satisfy y′′ 4y = 0 are 2x ± 2x − y1(x)= e and y2(x)= e− .

37.9. Theorem. Suppose the polynomial P (r) has n distinct roots r1,r2,...,rn. Then the general solution of L[y] = 0 is

r1x r2x rnx y = c1e + c2e + + cne where c1, c2,...,cn are arbitrary constants.

r1x r2x Proof. We have just seen that the functions y1(x) = e , y2(x) = e , y3(x) = er3x, etc. are solutions of the equation L[y] = 0. In Math 320 (or 319, or. . . ) you prove that these are all the solutions (it also follows from the method of variation of parameters that there aren’t any other solutions).

37.10. Complex roots and repeated roots. If the characteristic polynomial has n distinct real roots then Theorem 37.9 tells you what the general solution to the equation L[y] = 0 is. In general a polynomial equation like P (r) = 0 can have repeated roots, and it can have complex roots. 83

37.11. Example. Solve y′′ + 2y′ + y = 0. The characteristic polynomial is P (r) = r2 + 2r +1=(r + 1)2, so the only root of the characteristic equation r2 + 2r + 1 = 0 is r = 1 (it’s a repeated root). This means − x that for this equation we only get one exponential solution, namely y(x)= e− . It turns out that for this equation there is another solution which is not exponential. x It is y2(x)= xe− . You can check that it really satisﬁes the equation y′′ + 2y′ + y = 0. When there are repeated roots there are other solutions: if P (r) = 0, then tj ert is a solution if j is a nonnegative integer less than the multiplicity of r. Also, if any of the roots are complex, the phrase general solution should be understood to mean general complex solution and the coeﬃcients cj should be complex. If the equation is real, the real and imaginary part of a complex solution are again solutions. We only describe the case n = 2 in detail.

37.12. Theorem. Consider the diﬀerential equation d2y dy + a1 + a2y =0 ( ) dx2 dx † 2 and suppose that r1 and r2 are the solutions of the characteristic equation of r +a1r+a2 = 0. Then

(i): If r1 and r2 are distinct and real, the general solution of ( ) is † r1x r2x y = c1e + c2e .

(ii): If r1 = r2, the general solution of ( ) is † r1x r1x y = c1e + c2xe .

(iii): If r1 = α + βi and r2 = α βi, the general solution of ( ) is − † αx αx y = c1e cos(βx)+ c2e sin(βx).

In each case c1 and c2 are arbitrary constants. Case (i) and case (iii) can be subsumed into a single case using complex notation:

(α βi)x αx αx e ± = e cos βx ie sin βx, ±

(α+βi)x (α βi)x (α+βi)x (α βi)x e + e − e e − eαx cos βx = , eαx sin βx = − . 2 2i

38. Inhomogeneous Linear Equations

In this section we study the inhomogeneous linear diﬀerential equation

(n) (n 1) y + a1y − + + an 1y′ + any = k(x) − where the coeﬃcients a1,...,an are constants and the function k(x) is a given function. In the operator notation this equation may be written L[y]= k(x).

The following theorem says that once we know one particular solution yp of the inhomogeneous equation L[y]= k(x) we can ﬁnd all the solutions y to the inhomogeneous equation L[y]= k(x) by ﬁnding all the solutions yh to the homogeneous equation L[y] = 0. 84

38.1. Another Superposition Principle. Assume L[yp] = k(x). Then L[y] = k(x) if and only if y = yp + yh where L[yh] = 0.

Proof. Suppose L[yp]= k(x) and y = yp + yh. Then

L[y]= L[yp + yh]= L[yp]+ L[yh]= k(x)+ L[yh].

Hence L[y]= k(x) if and only if L[yh] = 0.

39. Variation of Constants

There is a method to ﬁnd the general solution of a linear inhomogeneous equation of arbitrary order, provided you already know the solutions to the homogeneous equation. We won’t explain this method here, but merely show you the answer you get in the case of second order equations.

If y1(x) and y2(x) are solutions to the homogeneous equation

y′′(x)+ a(x)y′(x)+ b(x)y(x) = 0 for which def W (x) = y1(x)y′ (x) y′ (x)y2(x) = 0, 2 − 1 then the general solution of the inhomogeneous equation

y′′(x)+ a(x)y′(x)+ b(x)y(x)= f(x) is given by y2(ξ)f(ξ) y1(ξ)f(ξ) y(x)= y1(x) dξ + y2(x) dξ. − W (ξ) W (ξ) For more details you should take a more advanced course like Math 319 or 320.

39.1. Undetermined Coefficients. The easiest way to find a particular solution yp to the inhomogeneous equation is the method of undetermined coefficients or “educated guessing.” Unlike the method of “variation of constants” which was (hardly) explained in the previous section, this method does not work for all equations. But it does give you the answer for a few equations which show up often enough to make it worth knowing the method. The basis of the “method” is this: it turns out that many of the second order equations with you run into have the form

y′′ + ay′ + by = f(t), where a and b are constants, and where the righthand side f(t) comes from a fairly short list of functions. For all f(t) in this list you memorize (yuck!) a particular solution yp. With the particular solution in hand you can then ﬁnd the general solution by adding it to the general solution of the homogeneous equation. Here is the list:

f(t)= polynomial in t: In this case you try yp(t) = some other polynomial in t with the same degree as f(t). Exceptions: if r = 0 is a root of the characteristic equation, then you must try a polynomial yp(t) of degree one higher than f(t); if r = 0 is a double root then the degree of yp(t) must be two more than the degree of f(t). at at f(t)= e : try yp(t)= Ae . Exceptions: if r = a is a root of the characteristic equation, then you must at try yp(t)= Ate ; 2 at if r = a is a double root then try yp(t)= At e . 85

f(t) = sin bt or f(t) = cos bt: In both cases, try yp(t)= A cos bt + B sin bt. Exceptions: if r = bi is a root of the characteristic equation, then you should try yp(t)= t(A cos bt + B sin bt). at at at f(t)= e sin bt or f(t)= e cos bt: Try yp(t)= e (A cos bt + B sin bt). Exceptions: if r = a + bi is a root of the characteristic equation, then you at should try yp(t)= te (A cos bt + B sin bt).

39.2. Example. Find the general solution to the following equations x (41) y′′ + xy′ y = 2e − (42) y′′ 2y′ + y = 1+ x2 − The first equation does not have constant coefficients so the method doesn’t apply. Sorry, but we can’t solve this equation in this course.7 The second equation does have constant coefficients, so we can solve the homogeneous equation (y′′ 2y′ + y = 0), but the righthand side does not appear in our list. Again, − the method doesn’t work.

39.3. A more upbeat example. To find a particular solution of 2 y′′ y′ + y = 3t − we note that (1) the equation is linear with constant coefficients, and (2) the right hand side is a polynomial, so it’s in our list of “right hand sides for which we know what to guess.” We try a polynomial of the same degree as the right hand side, namely 2. We don’t know which polynomial, so we leave its coefficients undetermined (whence the name of the method.) I.e. we try 2 yp(t)= A + Bt + Ct . To see if this is a solution, we compute

yp′ (t)= B + 2Ct, yp′′(t) = 2C, so that 2 yp′′ yp′ + yp =(A B + 2C)+(B 2C)t + Ct . −2 − − Thus y′′ y′ + yp = 3t if and only if p − p A B + 2C = 0, B 2C = 0, C = 3. − − Solving these equations leads to C = 3, B = 2C = 6 and A = B 2C = 0. We conclude − that 2 yp(t) = 6t + 3t is a particular solution.

39.4. Another example, which is rather long, but that’s because it is meant to cover several cases. Find the general solution to the equation 3 t 2t t y′′ + 3y′ + 2y = t + t e + 2e− e− sin 2t. − − Solution: First we ﬁnd the characteristic equation, r2 + 3r +2=(r + 2)(r + 1) = 0.

The characteristic roots are r1 = 1, and r2 = 2. The general solution to the homoge- − − neous equation is t 2t yh(t)= C1e− + C2e− .

7Who says you can’t solve this equation? For equation (41) you can ﬁnd a solution by computing its Taylor series! For more details you should again take a more advanced course (like Math 319), or, in this case, give it a try yourself. 86

We now look for a particular solutions. Initially it doesn’t look very good as the righthand side does not appear in our list. However, the righthand side is a sum of ﬁve terms, each of which is in our list.

Abbreviate L[y]= y′′ + 3y′ + 2y. Then we will find functions y1,...,y4 for which one has 3 t 2t t L[y1]= t + t , L[y2]= e , L[y3] = 2e− , L[y4]= e− sin 2t. − − def Then, by the Superposition Principle (Theorem 37.5) you get that yp = y1 + y2 + y3 + y4 satisfies 3 t 2t t L[yp]= L[y1]+ L[y2]+ L[y3]+ L[y4]= t + t e + 2e− e− sin 2t. − − So yp (once we find it) is a particular solution.

Now let’s ﬁnd y1,...,y4.

3 y1(t): the righthand side t + t is a polynomial, and r = 0 is not a root of the characteristic equation, so we try a polynomial of the same degree. Try 2 3 y1(t)= A + Bt + Ct + Dt . Here A,B,C,D are the undetermined coeﬃcients that give the method its name. You compute L [y1]= y1′′ + 3y1′ + 2y1 = (2C + 6Dt)+3(B + 2Ct + 3Dt2)+2(A + Bt + Ct2 + Dt3) = (2C + 3B + 2A)+(2B + 6C + 6D)t + (2C + 9D)t2 + 2Dt3.

3 So to get L[y1]= t + t we must impose the equations 2D = 1, 2C + 9D = 0, 2B + 6C + 6D = 1, 2C + 6B + 2A = 0. You can solve these equations one-by-one, with result D = 1 , C = 9 , B = 23 ,A = 87 , 2 − 4 − 4 8 and thus 87 23 9 2 1 3 y1(t)= t t + t . 8 − 4 − 4 2 t t at y2(t): We want y2(t) to satisfy L[y2]= e . Since e = e with a = 1, and a = 1 − t is not a characteristic root, we simply try y2(t)= Ae . A quick calculation gives t t t t L[y2]= Ae + 3Ae + 2Ae = 6Ae . t 1 To achieve L[y2]= e we therefore need 6A = 1, i.e. A = . Thus − − − 6 1 t y2(t)= e . − 6 2t 2t at y3(t): We want y3(t) to satisfy L[y3]= e− . Since e− = e with a = 2, and − 2t − a = 2 is a characteristic root, we can’t simply try y3(t)= Ae− . Instead you − 2t have to try y3(t)= Ate− . Another calculation gives 2t 2t 2t 2t L[y3]=(4t 4)Ae− + 3( 2t + 2)Ae− + 2Ate− (factor out Ae− ) − − 2t = (4 +3( 2) + 2)t +( 4+3) Ae− − − 2t = Ae− . − 2t Note that all the terms with te− cancel: this is no accident, but a consequence of the fact that a = 2 is a characteristic root. − 2t To get L[y3] = 2e− we see we have to choose A = 2. We ﬁnd − 2t y3(t)= 2te− . − 87

t y4(t): Finally, we need a function y4(t) for which one has L[y4]= e− sin 2t. The − list tells us to try t y4(t)= e− A cos 2t + B sin 2t . (Since 1 + 2i is not a root of the characteristic equation we are not in one of − the exceptional cases.) Diligent computation yields t t y4(t)= Ae− cos 2t + Be− sin 2t t t y4′ (t)= ( A + 2B)e− cos 2t + ( B 2A)e− sin 2t − t − − t y′′(t)= ( 3A 4B)e− cos 2t + ( 3B + 4A)e− sin 2t 4 − − − so that t t L[y4]=( 4A + 2B)e− cos 2t +( 2A 4B)e− sin 2t. − − − t We want this to equal e− sin 2t, so we have to find A, B with − 4A + 2B = 0, 2A 4B = 1. − − − − 1 The first equation implies B = 2A, the second then gives 10A = 1, so A = 10 2 − − and B = 10 . We have found 1 t 2 t − − y4(t)= 10 e cos 2t + 10 e sin 2t. After all these calculations we get the following impressive particular solution of our differential equation, 87 23 9 2 1 3 1 t 2t 1 t 2 t yp(t)= t t + t e 2te− + e− cos 2t + e− sin 2t 8 − 4 − 4 2 − 6 − 10 10 and the even more impressive general solution to the equation,

y(t)= yh(y)+ yp(t) t 2t = C1e− + C2e− + 87 23 t 9 t2 + 1 t3 8 − 4 − 4 2 1 t 2t 1 t 2 t e 2te− + e− cos 2t + e− sin 2t. − 6 − 10 10 You shouldn’t be put oﬀ by the fact that the result is a pretty long formula, and that the computations took up two pages. The approach is to (i) break up the right hand side into terms which are in the list at the beginning of this section, (ii) to compute the particular solutions for each of those terms and (iii) to use the Superposition Principle (Theorem 37.5) to add the pieces together, resulting in a particular solution for the whole right hand side you started with.

40. Applications of Second Order Linear Equations

40.1. Spring with a weight. In example 36.1 we showed that the height y(t) a mass m suspended from a spring with constant k satisﬁes k (43) my′′(t)+ ky(t)= mg, or y′′(t)+ y(t)= g. m k This is a Linear Inhomogeneous Equation whose homogeneous equation, y′′ + m y = 0 has

yh(t)= C1 cos ωt + C2 sin ωt as general solution, where ω = k/m. The right hand side is a constant, which is a polynomial of degree zero, so the method of “educated guessing” applies, and we can ﬁnd a particular solution by trying a constant yp = A as particular solution. You ﬁnd that k k mg yp′′ + m yp = m A, which will equal g if A = k . Hence the general solution to the “spring with weight equation” is mg y(t)= + C cos ωt + C sin ωt. k 1 2 88

To solve the initial value problem y(0) = y0 and y′(0) = v0 we solve for the constants C1 and C2 and get mg v0 mg y(t)= + sin(ωt)+ y0 cos(ωt). k ω − k which you could rewrite as mg y(t)= + Y cos(ωt φ) k − for certain numbers Y,φ. The weight in this model just oscillates up and down forever: this motion is called a simple harmonic oscillation, and the equation (43) is called the equation of the Harmonic Oscillator.

40.2. The pendulum equation. In example 36.2 we saw that the angle θ(t) sub- tended by a swinging pendulum satisfies the pendulum equation, d2θ g (38) + sin θ(t) = 0. dt2 L This equation is not linear and cannot be solved by the methods you have learned in this course. However, if the oscillations of the pendulum are small, i.e. if θ is small, then we can approximate sin θ by θ. Remember that the error in this approximation is the remainder term in the Taylor expansion of sin θ at θ = 0. According to Lagrange this is θ3 sin θ = θ + R2(θ), R2(θ) = cos θ˜ with θ˜ θ. 3! | |≤ When θ is small, e.g. if θ 10◦ 0.175 radians then compared to θ the error is at most | |≤ ≈ R (θ) (0.175)2 3 0.005, θ ≤ 3! ≈ in other words, the error is no more than half a percent. So for small angles we will assume that sin θ θ and hence θ(t) almost satisfies the ≈ equation d2θ g (44) + θ(t) = 0. dt2 L In contrast to the pendulum equation (38), this equation is linear, and we could solve it right now. The procedure of replacing inconvenient quantities like sin θ by more manageable ones (like θ) in order to end up with linear equations is called linearization. Note that the solutions to the linearized equation (44), which we will derive in a moment, are not solutions of the Pendulum Equation (38). However, if the solutions we find have small angles (have θ small), then the Pendulum Equation and its linearized form (44) are | | almost the same, and “you would think that their solutions should also be almost the same.” I put that in quotation marks, because (1) it’s not a very precise statement and (2) if it were more precise, you would have to prove it, which is not easy, and not a topic for this course (or even Math 319 – take Math 419 or 519 for more details.) Let’s solve the linearized equation (44). Setting θ = ert you find the characteristic equation g r2 + = 0 L which has two complex roots, r = i g . Therefore, the general solution to (44) is ± ± L g g θ(t)= A cos t + B sin t , L L 89 and you would expect the general solution of the Pendulum Equation (38) to be almost the same. So you see that a pendulum will oscillate, and that the period of its oscillation is given by L T = 2π . g Once again: because we have used a linearization, you should expect this statement to be valid only for small oscillations. When you study the Pendulum Equation instead of its linearization (44), you discover that the period T of oscillation actually depends on the amplitude of the oscillation: the bigger the swings, the longer they take.

40.3. The eﬀect of friction. A real weight suspended from a real spring will of course not oscillate forever. Various kinds of friction will slow it down and bring it to a stop. As an example let’s assume that air drag is noticeable, so, as the weight moves the surrounding air will exert a force on the weight (To make this more likely, assume the weight is actually moving in some viscous liquid like salad oil.) This drag is stronger as the weight moves faster. A simple model is to assume that the friction force is proportional to the velocity of the weight,

Ffriction = hy′(t). − y(t) This adds an extra term to the oscillator equation (43), and gives

my′′(t)= Fgrav + Ffriction = ky(t)+ mg hy′(t) F F − − drag spring i.e. m (45) my′′(t)+ hy′(t)+ ky(t)= mg. Fgravity This is a second order linear homogeneous differential equation with constant coefficients. salad dressing A particular solution is easy to find, yp = mg/k works again. To solve the homogeneous equation you try y = ert, which leads to the characteristic equation mr2 + hr + k = 0, whose roots are h √h2 4mk r = − ± − ± 2m If friction is large, i.e. if h> √4km, then the two roots r are real, and all solutions are ± of exponential type,

mg r+t r−t y(t)= + C+e + C e . k − Both roots r are negative, so all solutions satisfy ± lim y(t) = 0. t →∞

If friction is weak, more precisely, if h < √4mk then the two roots r are complex ± numbers, h √4km h2 r = iω, with ω = − . ± − 2m ± 2m The general solution in this case is

h mg t y(t)= + e− 2m A cos ωt + B sin ωt . k These solutions also tend to zero as t , but they oscillate inﬁnitely often. →∞ 90

40.4. Electric circuits. Many equations in physics and engineering have the form (45). For example in the electric circuit in the diagram a time varying voltage Vin(t) is applied to a resistor R, an inductance L and a capacitor C. This causes a current I(t) to ﬂow through the circuit. How much is this current, and how much is, say, the voltage across the resistor?

V C

C V L L

V in

V = V R R out

Electrical engineers will tell you that the total voltage Vin(t) must equal the sum of the voltages VR(t), VL(t) and VC (t) across the three components. These voltages are related to the current I(t) which ﬂows through the three components as follows:

VR(t)= RI(t) dV (t) 1 C = I(t) dt C dI(t) V (t)= L . L dt Surprisingly, these little electrical components know calculus! (Here R, C and L are constants depending on the particular components in the circuit. They are measured in “Ohm,” “Farad,” and “Henry.”) Starting from the equation

Vin(t)= VR(t)+ VL(t)+ VC (t) you get

Vin′ (t)= VR′ (t)+ VL′ (t)+ VC′ (t) 1 = RI′(t)+ LI′′(t)+ I(t) C In other words, for a given input voltage the current I(t) satisﬁes a second order inhomogeneous linear diﬀerential equation

d2I dI 1 (46) L + R + I = V ′ (t). dt2 dt C in

Once you know the current I(t) you get the output voltage Vout(t) from

Vout(t)= RI(t).

In general you can write down a diﬀerential equation for any electrical circuit. As you add more components the equation gets more complicated, but if you stick to resistors, inductances and capacitors the equations will always be linear, albeit of very high order. 91

41. PROBLEMS

GENERAL QUESTIONS 332. Classify each of the following as homogeneous linear, inhomogeneous linear, or nonlinear and specify the order. For each linear equation say whether or not the coeﬃcients are constant.

(i) y′′ + y =0 (ii) xy′′ + yy′ = 0

(iii) xy′′ y′ =0 (iv) xy′′ + yy′ = x − x (v) xy′′ y′ = x (vi) y′ + y = xe . − 2 333. (i) Show that y = x + 5 is a solution of xy′′ y′ = 0. − 2 (ii) Show that y = C1x + C2 is a solution of xy′′ y′ = 0. − 2 334. (i) Show that y = (tan(c1x + c2))/c1 is a solution of yy′′ = 2(y′) 2y′. − (ii) Show that y1 = tan(x) and y2 = 1 are solutions of this equation, but that y1 + y2 is not. (iii) Is the equation linear homogeneous? SEPARATION OF VARIABLES

dy 335. Consider the differential equation 337. +(1+3x2)y = 0,y(1) = 1. dx dy 4 y2 = − . dy dt 4 338. + x cos2 y = 0,y(0) = π . dx 3 (a) Find the solutions y0, y1, y2, and dy 1+ x y which satisfy y (0) = 0, y (0) = 1, 339. + = 0,y(0) = A. 3 0 1 dx 1+ y y2(0) = 2 and y3(0) = 3. dy 2 (b) Find limt yk(t) for k = 340. + 1 y = 0,y(0) = A. →∞ dx − 1, 2, 3. dy 2 (c) Find limt yk(t) for k = 341. +1+ y = 0,y(0) = A. →−∞ dx 1, 2, 3. 342. Find the function y of x which satis- (d) Graph the four solutions y0,..., fies the initial value problem: y3. dy x2 1 (e) Show that the quantity x = + − = 0 y(0) = 1 dx y (y + 2)/4 satisfies the so-called Logistic Equation 343. Find the general solution of dx dy = x(1 x). + 2y + ex 0 dt − dx ≡ (Hint: if x =(y + 2)/4, then y = 4x 2; − dy substitute this in both sides of the diffeq 344. (cos x)y = esin x,y(0) = A. for y). dx − dy 345. y2 + x3 = 0,y(0) = A. dx ∗ ∗ ∗ 346. Group problem. Read Example In each of the following problems you 35.2 on “Leaky bucket dating” again. In should find the function y of x which sat- that example we assumed that a √K = isfies the conditions (A is an unspecified A 2. constant: you should at least indicate for which values of A your solution is valid.) (a) Solve diffeq for h(t) without assuming a √K = 2. Abbreviate C = dy 2 A 336. + x y = 0,y(1) = 5. a √K. dx A 92

(b) If in an experiment one found that the bucket empties in 20 seconds after being ﬁlled to height 20 cm, then how much is the constant C?

LINEAR HOMOGENEOUS

3 347. (a) Show that y = 4ex + 7e2x is a so- d y 357. 3 y = 0 lution of y′′ 3y′ + 2y = 0. dx − − x 2x (4) (b) Show that y = C1e + C2e is a so- 358. y (t) 2y′′(t) 3y(t) = 0 lution of y′′ 3y′ + 2y = 0. − − − (4) (c) Find a solution of y′′ 3y′ + 2y = 0 359. y (t) + 4y′′(t) + 3y(t) = 0. − ′ (4) such that y(0) = 7 and y (0) = 9. 360. y (t) + 2y′′(t) + 2y(t) = 0. dy 348. (a) Find all solutions of + 2y = 0. (4) dx 361. y (t)+ y′′(t) 6y(t) = 0. dy x (b) Find all solutions of + 2y = e− . − dx (4) dy x 362. y (t) 8y′′(t) + 15y(t) = 0. (c) Find y if dx + 2y = e− and y = 7 − when x = 0. 363. f ′′′(x) 125f(x) = 0. − 349. (a) Find all real solutions of 364. u(5)(x) 32u(x) = 0. 2 − d y dy (5) 6 + 10y = 0. 365. u (x) + 32u(x) = 0. dt2 − dt 366. y′′′(t) 5y′′(t) + 6y′(t) 2y(t) = 0. (b) Find y if − − (4) (3) 367. h (t) h (t) + 4h′′(t) 4h(t) = 0. y′′ 6y′ + 10y = 0, − − − 368. z′′′(x) 5z′′(x) + 4z(x) = 0. and in addition y satisfies the initial con- − ditions y(0) = 7, and y′(0) = 11. 350. Solve the initial value problem: ∗ ∗ ∗ Solve each of the following initial value y′′ 5y′ + 4y 0 − ≡ problems. Your final answer should not y(0) = 2 use complex numbers, but you may use y′(0) = 1 complex numbers to find it. − 369. y′′ + 9y = 0, y(0) = 0,y′(0) = 3. 351. For y as a function of x, find the gen- − eral solution of the equation: 370. y′′ + 9y = 0, y(0) = 3,y′(0) = 0. − y′′ 2y′ + 10y 0 371. y′′ 5y′+6y = 0, y(0) = 0,y′(0) = 1. − ≡ − 372. y′′+5y′+6y = 0, y(0) = 1,y′(0) = 0.

∗ ∗ ∗ 373. y′′+5y′+6y = 0, y(0) = 0,y′(0) = 1. Find the general solution y = y(x) of the 374. y′′ 6y′+5y = 0, y(0) = 1,y′(0) = 0. following diﬀerential equations − 4 375. y′′ 6y′+5y = 0, y(0) = 0,y′(0) = 1. d y − 352. 4 = y dx 376. y′′+6y′+5y = 0, y(0) = 1,y′(0) = 0. d4y 353. + y = 0 377. y′′+6y′+5y = 0, y(0) = 0,y′(0) = 1. dx4 378. y′′ 4y′+5y = 0, y(0) = 1,y′(0) = 0. d4y d2y − 354. = 0 dx4 − dx2 379. y′′ 4y′+5y = 0, y(0) = 0,y′(0) = 1. − 4 2 d y d y 380. y′′+4y′+5y = 0, y(0) = 1,y′(0) = 0. 355. + = 0 dx4 dx2 381. y′′+4y′+5y = 0, y(0) = 0,y′(0) = 1. d3y 356. + y = 0 382. y′′ 5y′+6y = 0, y(0) = 1,y′(0) = 0. dx3 − 93

383. f ′′′(t)+ f ′′(t) f ′(t) + 15f(t) = 0, − with initial conditions f(0) = 0, f ′(0) = 1, f ′′(0) = 0.

LINEAR INHOMOGENEOUS

2 384. Find particular solutions of d y 386. 2 y = 2 3x dt − y′′ 3y′ + 2y = e − 2 x d y t y′′ 3y′ + 2y = e 387. y = 2e − dt2 − 3x x y′′ 3y′ + 2y = 4e + 5e d2y − 388. + 9y = cos 3t dt2 ∗ ∗ ∗ 2 385. Find a particular solution of the d y 389. 2 + 9y = cos t equation: dt 2 x d y y′′ + y′ + 2y = e + x + 1 390. + y = cos t dt2 Find the general solution y(t) of the d2y 391. + y = cos 3t. following diﬀerential equations dt2

392. Find y if d2y dy (a) + 2 + y = 0 y(0) = 2, y′(0) = 3 dx2 dx 2 d y dy x (b) + 2 + y = e− y(0) = 0, y′(0) = 0 dx2 dx 2 d y dy x (c) + 2 + y = xe− y(0) = 0, y′(0) = 0 dx2 dx 2 d y dy x x (d) + 2 + y = e− + xe− y(0) = 2, y′(0) = 3. dx2 dx Hint: Use the Superposition Principle to save work. 393. Group problem. (i) Find the general solution of it z′′ + 4z′ + 5z = e using complex exponentials. (ii) Solve z′′ + 4z′ + 5z = sin t using your solution to question (i). (iii) Find a solution for the equation

(1 i)t z′′ + 2z′ + 2z = 2e− − (1 i)t in the form z(t)= u(t)e− − . (iv) Find a solution for the equation t x′′ + 2x′ + 2x = 2e− cos t. Hint: Take the real part of the previous answer. (v) Find a solution for the equation t y′′ + 2y′ + 2y = 2e− sin t. 94

APPLICATIONS

394. A population of bacteria grows at a rate proportional to its size. Write and solve a differential equation which expresses this. If there are 1000 bacteria after one hour and 2000 bacteria after two hours, how many bacteria are there after three hours? 395. Rabbits in Madison have a birth rate of 5% per year and a death rate (from old age) of 2% per year. Each year 1000 rabbits get run over and 700 rabbits move in from Sun Prairie. (i) Write a differential equation which describes Madison’s rabbit population at time t. (ii) If there were 12,000 rabbits in Madison in 1991, how many are there in 1994? 396. Group problem. According to Newton’s law of cooling the rate dT/dt at which an object cools is proportional to the difference T A between its temperature T and the − ambient temperature A. The differential equation which expresses this is dT = k(T A) dt − where k< 0 and A are constants. (i) Solve this equation and show that every solution satisfies lim T = A. t →∞ (ii) A cup of coffee at a temperature of 180oF sits in a room whose temperature is 750F. In five minutes its temperature has dropped to 150◦F. When will its temperature be 90◦F? What is the limit of the temperature as t ? →∞ 397. Retaw is a mysterious living liquid; it grows at a rate of 5% of its volume per hour. A scientist has a tank initially holding y0 gallons of retaw and removes retaw from the tank continuously at the rate of 3 gallons per hour. (i) Find a differential equation for the number y(t) of gallons of retaw in the tank at time t.

(ii) Solve this equation for y as a function of t. (The initial volume y0 will appear in your answer.)

(iii) What is limt y(t) if y0 = 100? →∞ (iv) What should the value of y0 be so that y(t) remains constant? 398. A 1000 gallon vat is full of 25% solution of acid. Starting at time t = 0 a 40% solution of acid is pumped into the vat at 20 gallons per minute. The solution is kept well mixed and drawn oﬀ at 20 gallons per minute so as to maintain the total value of 1000 gallons. Derive an expression for the acid concentration at times t> 0. As t what percentage →∞ solution is approached? 399. The volume of a lake is V = 109 cubic feet. Pollution P runs into the lake at 3 cubic feet per minute, and clean water runs in at 21 cubic feet per minute. The lake drains at a rate of 24 cubic feet per minute so its volume is constant. Let C be the concentration of pollution in the lake; i.e. C = P/V . (i) Give a diﬀerential equation for C.

(ii) Solve the diﬀerential equation. Use the initial condition C = C0 when t = 0 to evaluate the constant of integration.

(iii) There is a critical value C∗ with the property that for any solution C = C(t) we have

lim C = C∗. t →∞ Find C∗. If C0 = C∗, what is C(t)? 95

400. Group problem. A philanthropist endows a chair. This means that she donates an amount of money B0 to the university. The university invests the money (it earns interest) and pays the salary of a professor. Denote the interest rate on the investment by r (e.g. if r = .06, then the investment earns interest at a rate of 6% per year) the salary of the professor by a (e.g. a = $50, 000 per year), and the balance in the investment account at time t by B. (i) Give a diﬀerential equation for B.

(ii) Solve the diﬀerential equation. Use the initial condition B = B0 when t = 0 to evaluate the constant of integration.

(iii) There is a critical value B∗ with the property that (1) if B0 < B∗, then there is a t > 0 with B(t) = 0 (i.e. the account runs out of money) while (2) if B0 > B∗, then limt B = . Find B∗. →∞ ∞ (iv) This problem is like the pollution problem except for the signs of r and a. Explain.

401. Group problem. A citizen pays social security taxes of a dollars per year for T1 years, then retires, then receives payments of b dollars per year for T2 years, then dies. The account which receives and dispenses the money earns interest at a rate of r% per year and has no money at time t = 0 and no money at the time t = T1 +T2 of death. Find two diﬀerential equations for the balance B(t) at time t; one valid for 0 t T1, the other ≤ ≤ valid for T1 t T1 + T2. Express the ratio b/a in terms of T1, T2, and r. Reasonable ≤ ≤ values for T1, T2, and r are T1 = 40, T2 = 20, and r = 5% = .05. This model ignores inﬂation. Notice that 0 < dB/dt for 0

402. A 300 gallon tank is full of milk containing 2% butterfat. Milk containing 1% butterfat is pumped in a 10 gallons per minute starting at 10:00 AM and the well mixed milk is drained oﬀ at 15 gallons per minute. What is the percent butterfat in the milk in the tank 5 minutes later at 10:05 AM? Hint: How much milk is in the tank at time t? How much butterfat is in the milk at time t = 0?

403. A sixteen pound weight is suspended from the lower end of a spring whose upper end is attached to a rigid support. The weight extends the spring by half a foot. It is struck by a sharp blow which gives it an initial downward velocity of eight feet per second. Find its position as a function of time.

404. A sixteen pound weight is suspended from the lower end of a spring whose upper end is attached to a rigid support. The weight extends the spring by half a foot. The weight is pulled down one feet and released. Find its position as a function of time.

405. The equation for the displacement y(t) from equilibrium of a spring subject to a forced vibration of frequency ω is d2y (47) + 4y = sin(ωt). dt2

(i) Find the solution y = y(ω,t) of (47) for ω = 2 if y(0) = 0 and y′(0) = 0. (ii) What is limω 2 y(ω,t)? → (iii) Find the solution y(t) of d2y (48) + 4y = sin(2t) dt2

if y(0) = 0 and y′(0) = 0. (Hint: Compare with (47).) 96

406. Group problem. Suppose that an undamped spring is subjected to an external periodic force so that its position y at time t satisﬁes the diﬀerential equation d2y + ω2y = c sin(ωt). dt2 0 (i) Show that the general solution is c y = C cos ω t + C sin ω t + sin ωt. 1 0 2 0 ω2 ω2 0 − when ω0 = ω. (ii) Solve the equation when ω = ω0. (iii) Show that in part (i) the solution remains bounded as t but in part (ii) this → ∞ is not so. (This phenomenon is called resonance. To see an example of resonance try Googling “Tacoma Bridge Disaster.”) 407. Group problem. Have look at the electrical circuit equation (46) from 40.4. § (i) Find the general solution of (46), assuming that Vin(t) does not depend on time t. What is limt I(t)? →∞ (ii) Assume for simplicity that L = C = 1, and that the resistor has been short circuited, i.e. that R = 0. If the input voltage is a sinusoidal wave,

Vin(t)= A sin ωt, (ω = 1) then ﬁnd a particular solution, and then the general solution. (iii) Repeat problem (ii) with ω = 1. (iv) Suppose again that L = C = 1, but now assume that R > 0. Find the general solution when Vin(t) is constant. (v) Still assuming L = C = 1, R > 0 ﬁnd a particular solution of the equation when the input voltage is a sinusoidal wave

Vin(t)= A sin ωt.

408. You are watching a buoy bobbing up and down in the water. Assume that the buoy height with respect to the surface level of the water satisfies the damped oscillator equation: z′′ + bz′ + kz 0 where b and k are positive constants. ≡ Something has initially disturbed the buoy causing it to go up and down, but friction will gradually cause its motion to die out. You make the following observations: At time zero the center of the buoy is at z(0) = 0 , i.e., the position it would be in if it were at rest. It then rises up to a peak and falls down so that at time t = 2 it again at zero, z(2) = 0 descends downward and then comes back to 0 at time 4, i.e, z(4) = 0. Suppose z(1) = 25 and z(3) = 16. − (a) How high will z be at time t = 5? (b) What are b and k? Hint: Use that z = Aeαt sin(ωt + B). 409. Contrary to what one may think the buoy does not reach its peak at time t = 1 which is t midway between its first two zeros, at t = 0 and t = 2. For example, suppose z = e− sin t. π Then z is zero at both t = 0 and t = π. Does z have a local maximum at t = 2 ? 410. In the buoy problem 408 suppose you make the following observations: It rises up to its first peak at t = 1 where z(1) = 25 and then descends downward to a local minimum at at t = 3 where z(3) = 16. − (a) When will the buoy reach its second peak and how high will that be? (b) What are b and k? 97

Note: It will not be the case that z(0) = 0.

Linear operators

2 n Given a polynomial p = p(x)= a0 + a1x + a2x + . . . + anx and z = z(t) an infinitely differentiable function of t define

(1) (2) (n) Lp(z)= a0z + a1z + a2z + . . . + anz

k (k) d z th where z = dtk is the k derivative of z with respect to t.

411. Show for L = Lp that

(a) L(z1 + z2)= L(z1)+ L(z2) (b) L(Cz)= CL(z) where C is any constant Such an L is called a linear operator. Operator because it takes as input a function and then ouputs another function. Linear refers to properties (a) and (b).

rt rt 412. Let r be any constant and p any polynomial. Show that Lp(e )= p(r)e .

413. For p and q polynomials show that

Lp+q(z)= Lp(z)+ Lq(z)

414. For p and q polynomials show that

Lp q (z)= Lp(Lq(z)) Here p q refers to the ordinary product of the two polynomials. 415. Let α be a constant. For any u an inﬁnitely diﬀerentiable function of t show that αt (1) αt (a) Lx α(u e )= u e − αt (n) αt (b) L(x α)n (u e )= u e − 416. Let α be any constant, p a polynomial, and suppose that (x α)n divides p. Show that − for any k < n k αt Lp(t e ) 0 ≡

n1 nm 417. Suppose that p(x)=(x α1) (x αm) where the αi are distinct constants. − − Suppose that

1 α t 2 α t n n 1 α t 1 αmt 2 αmt nm nm 1 αmt z = C e 1 + C te 1 + C 1 t 1− e 1 + + C e + C te + C t − e 1 1 1 m m m Show that Lp(z) 0. ≡

In a more advanced course in the theory of diﬀerential equations it would be proved that every solution of Lp(z) 0 has this form, i.e., z satisﬁes the above formula for some i ≡ choice of the constants Cj .

418. Suppose L is a linear operator and b = b(t) is a ﬁxed function of t. Suppose that zP is one particular solution of L(z)= b, i.e., L(zP )= b. Suppose that z is any other solution of L(z)= b. Show that L(z zP ) 0. Show that for any solution of the equation L(z)= b − ≡ there is a solution zH of the associated homogenous equation such that z = zP + zH .

Variations of Parameters 98

419. Given the equation L(z)= z′′ + a0z′ + a1z b ≡ where a0,a1,b are given functions of t. Then

L(fz1 + gz2)= b where zh = C1z1 + C2z2 is the general solution of the associated homogenous equation L(z) 0 and the derivatives ≡ of f and g satisfy: 0 z z 0 det 2 det 1 b z2′ z1′ b f ′ = g′ = z z z z det 1 2 det 1 2 z′ z′ z′ z′ 1 2 1 2 Use these formulas to ﬁnd the general solution of 2 z′′ + z (cos t) ≡ 420. Use these formulas to ﬁnd the general solution of 1 z′′ + z ≡ sin t 421. Solve the initial value problem: 2 z′′ + z (tan t) ≡ z(0) = 1 z′(0) = 1 − 422. Find the general solution of 1 ′′ z z t t − ≡ e + e− 423. Given a system of linear equations ax + by = r cx + dy = s show that the solution is given by:

r b a r det det s d c s x = y = a b a b det det c d c d You may assume the determinant: a b det = ad bc c d − is not zero.

424. Given the linear operator L(z)= z′′ + a0z′ + a1z suppose L(z1) 0 and L(z2) 0 and ≡ ≡ that f and g are functions of t which satisfy f ′z1 + g′z2 0. Show that ≡ L(fz1 + gz2)= f ′z1′ + g′z2′

425. Prove that the formulas given problem 419 work. 99

Chapter 5: Vectors

42. Introduction to vectors

42.1. Deﬁnition. A vector is a column of two, three, or more numbers, written as

a1 a1 a1 . a = or a = a2 or a = . a2    .  a3 a  n     in general.

a1 The length of a vector a = a2 is deﬁned by a3 a1 2 2 2 a = a2 = a + a + a .   1 2 3 a3   We will always deal with either the two or three dimensional cases, in other words, the cases n = 2 or n = 3, respectively. For these cases there is a geometric description of vectors which is very useful. In fact, the two and three dimensional theories have their origins in mechanics and geometry. In higher dimensions the geometric description fails, simply because we cannot visualize a four dimensional space, let alone a higher dimensional space. Instead of a geometric description of vectors there is an abstract theory called Linear Algebra which deals with “vector spaces” of any dimension (even inﬁnite!). This theory of vectors in higher dimensional spaces is very useful in science, engineering and economics. You can learn about it in courses like math 320 or 340/341.

42.2. Basic arithmetic of vectors. You can add and subtract vectors, and you can multiply them with arbitrary real numbers. this section tells you how. The sum of two vectors is deﬁned by a b a + b (49) 1 + 1 = 1 1 , a2 b2 a2 + b2 and a1 b1 a1 + b1 a2 + b2 = a2 + b2 .       a3 b3 a3 + b3 The zero vector is deﬁned by      0 0 0 = or 0 = 0 . 0   0 It has the property that   a + 0 = 0 + a = a no matter what the vector a is. 100

a1 You can multiply a vector a = a2 with a real number t according to the rule   a3   ta1 ta = ta2 .   ta3   In particular, “minus a vector” is deﬁned by

a1 − a =( 1)a = a2 . − − −  a3 −   The diﬀerence of two vectors is deﬁned by

a b = a +( b). − − So, to subtract two vectors you subtract their components,

a1 b1 a1 b1 − a b = a2 b2 = a2 b2 −   −    −  a3 b3 a3 b3 −       42.3. Some GOOD examples. 2 3 1 1 0 2 + − = − 2 + 3 = 3 π 3+ π 0 1 3 1 1 2 1 0 0 a −12 12 b 0 = − a 0 + b 1 + c 0 = 3 − √2 3 √2 0 0 1 c − 12√39 0 t + t2 t 0 = = 0 = (1+ t) π2 ln 3 0 1 t2 1 t − − −

42.4. Two very, very BAD examples. Vectors must have the same size to be added, therefore 1 2 + 3 = undeﬁned!!! 3   2 Vectors and numbers are diﬀerent things,  so an equation like a = 3 is nonsense! This equation says that some vector (a) is equal to some number (in this case: 3). Vectors and numbers are never equal!

42.5. Algebraic properties of vector addition and multiplication. Addition of vectors and multiplication of numbers and vectors were deﬁned in such a way that the following always hold for any vectors a, b, c (of the same size) and any real numbers s,t

(50) a + b = b + a [vector addition is commutative] (51) a +(b + c)=(a + b)+ c [vector addition is associative] (52) t(a + b)= ta + tb [ﬁrst distributive property] (53) (s + t)a = sa + ta [second distributive property] 101

a1 b1 42.6. Prove (50). Let a = a2 and b = b2 be two vectors, and consider both a 3 b3 possible ways of adding them:

a1 b1 a1 + b1 b1 a1 b1 + a1 a2 + b2 = a2 + b2 and b2 + a2 = b2 + a2             a3 b3 a3 + b3 b3 a3 b3 + a3             We know (or we have assumed long ago) that addition of real numbers is commutative, so that a1 + b1 = b1 + a1, etc. Therefore

a1+b1 b1+a1 a + b = a2+b2 = b2+a2 = b + a. a +b b +a 3 3 3 3 This proves (50).

42.7. Example. If v and w are two vectors, we deﬁne

a = 2v + 3w , b = v + w . −

Problem: Compute a + b and 2a 3b in terms of v and w . − Solution:

a + b = (2v + 3w )+( v + w )=(2 1)v +(3+1)w = v + 4w − − 2a 3b = 2(2v + 3w ) 3( v + w ) = 4w + 6w + 3v 3w = 7v + 3w . − − − −

Problem: Find s,t so that sa + tb = v.

Solution: Simplifying sa + tb you ﬁnd

sa + tb = s(2v + 3w )+ t( v + w )=(2s t)v + (3s + t)w . − − One way to ensure that sa+ tb = v holds is therefore to choose s and t to be the solutions of

2s t = 1 − 3s + t = 0

The second equation says t = 3s. The ﬁrst equation then leads to 2s+3s = 1, i.e. s = 1 . − 5 Since t = 3s we get t = 3 . The solution we have found is therefore − − 5 1 a 3b = v. 5 − 5

42.8. Geometric description of vectors. Vectors originally appeared in mechanics, where they represented forces: a force acting on some object has a magnitude and a direction. Thus a force can be thought of as an arrow, where the length of the arrow indicates how strong the force is (how hard it pushes or pulls). So we will think of vectors as arrows: if you specify two points P and Q, then the arrow pointing from P to Q is a vector and we denote this vector by −−→PQ. The precise mathematical deﬁnition is as follows: −−→PQ 102 P Q p q 42.9. Deﬁnition. For any pair of points P and Q whose coordinates are (p1,p2,p3)

and (q1, q2, q3) one deﬁnes a vector −−→PQ by Q q1 p1 2

− p −−→PQ = q2 p2 . −−→PQ

 −  −

q3 p3 2

− q   P If the initial point of an arrow is the origin O, and the ﬁnal point is any point Q, then the q1 p1 vector −−→OQ is called the position vector of the point Q. − two pictures of If p and q are the position vectors of P and Q, then one can write −−→PQ as the vector −−→PQ = q p A − a 2 q1 p1 a −−→PQ = q2 p2 = q p.   −   − q3 p3     O a1 q p 1− 1 For plane vectors we deﬁne −−→PQ similarly, namely, −−→PQ = q2 p2 . The old formula for the distance between two points P and Q in the plane − a3 A 2 2 distance from P to Q = (q1 p1) +(q2 p2) − − a says that the length of the vector −−→PQ is just the distance between the points P and Q, a2 i.e. O distance from P to Q = −−→PQ . a1 This formula is also valid if P and Q are points in space. position vectors in the plane and in space 42.10. Example. The point P has coordinates (2, 3); the point Q has coordinates (8, 6). The vector −−→PQ is therefore

8 2 6 −−→PQ = − = . 6 3 3 − This vector is the position vector of the point R whose coordinates are (6, 3). Thus

6 Q 6 5 −−→PQ = −→OR = . −−→PQ 3 4 3 P The distance from P to Q is the length of the vector R 2 −−→PQ, i.e. 1 −→OR 6 distance P to Q = = 62 + 32 = 3√5. O 1 2 3 4 5 6 7 8 3 42.11. Example. Find the distance between the points A and B whose position 1 0 vectors are a = 1 and b = 1 respectively. 0 1 Solution: One has 1 − distance A to B = −→AB = b a = 0 = ( 1)2 + 02 + 12 = √2 −   − 1   103

42.12. Geometric interpretation of vector addition and multiplication. Sup- pose you have two vectors a and b. Consider them as position vectors, i.e. represent them by vectors that have the origin as initial point: a = −→OA, b = −−→OB. Then the origin and the three endpoints of the vectors a, b and a+b form a parallelogram. See ﬁgure 15. To multiply a vector a with a real number t you multiply its length with t ; if t< 0 | | you reverse the direction of a.

a a + b a + b y b

a b a a + b b

Figure 15. Two ways of adding plane vectors, and an addition of space vectors

a b − a 2a

a b − b

a − a b a − − Figure 16. Multiples of a vector, and the diﬀerence of two vectors.

2v a + b = v + 4w

w 3 a a a w w w 3 b b w 2v −v v v

Figure 17. Picture proof that a + b = v + 4w in example 42.13. 104

42.13. Example. In example 42.7 we assumed two vectors v and w were given, and then defined a = 2v + 3w and b = v + w . In figure 17 the vectors a and b − are constructed geometrically from some arbitrarily chosen v and w . We also found algebraically in example 42.7 that a + b = v + 4w . The third drawing in figure 17 illustrates this.

43. Parametric equations for lines and planes

B Given two distinct points A and B we consider the line segment AB. If X is any given point on AB then we will now ﬁnd a formula −−→AX for the position vector of X. X −→AB Deﬁne t to be the ratio between the lengths of the line segments A AX and AB, length AX t = . length AB Then the vectors −−→AX and −→AB are related by −−→AX = t−→AB. Since AX is shorter than AB we have 0

−−→OX = −→OA + −−→AX = −→OA + t−→AB. If we write a,b, x for the position vectors of A,B,X, then we get (54) x = (1 t)a + tb = a + t(b a). − − This equation is called the parametric equation for the line through A and B. In our derivation the parameter t satisﬁed 0 t 1, but there is nothing that keeps us from ≤ ≤ substituting negative values of t, or numbers t > 1 in (54). The resulting vectors x are position vectors of points X which lie on the line ℓ through A and B.

ℓ ℓ

a B X B b − = a) −→AB t(b − b b x A A a a

O O

Figure 18. Constructing points on the line through A and B

43.1. Example. [Find the parametric equation for the line ℓ through the points A(1, 2) and B(3, 1), and determine where ℓ intersects the x1 axis. ] − 105

1 x2 Solution: The position vectors of A, B are a = ( 2 ) and b = 3 , so the position vector of an arbitrary point on ℓ is 3 1 given by− A 2 1 3 1 1 2 1 + 2t x = a + t(b a)= + t − = + t = − 2 1 2 2 3 2 3t 1 − − − − where t is an arbitrary real number. 1 2 3 x1 This vector points to the point X = (1+2t, 2 3t). By − B deﬁnition, a point lies on the x1-axis if its x2 component vanishes. Thus if the point X =(1+2t, 2 3t) − 2 lies on the x1-axis, then 2 3t = 0, i.e. t = 3 . The intersection point X of ℓ and the − 2 5 x1-axis is therefore X =(1+2 , 0) =( , 0). |t=2/3 3 3 43.2. Midpoint of a line segment. If M is the midpoint of the line segment AB, then the vectors −−→AM and −−→MB are both parallel and have the same direction and length (namely, half the length of the line segment AB). Hence they are equal: −−→AM = −−→MB. If a, m , and b are the position vectors of A , M and B, then this means m a = −−→AM = −−→MB = b m . − − Add m and a to both sides, and divide by 2 to get a + b m = 1 a + 1b = . 2 2 2 43.3. Parametric equations for planes in space*. You can specify a plane in three dimensional space by naming a point A on the plane P, and two vectors v and w parallel to the plane P, but not parallel to each other. Then any point on the plane P has position vector x given by (55) x = a + sv + tw .

x = a a + w sv + A tw

v X sv P tw

Figure 19. Generating points on a plane P

The following construction explains why (55) will give you any point on the plane through A, parallel to v, w . Let A, v, w be given, and suppose we want to express the position vector of some other point X on the plane P in terms of a = −→OA, v, and w . 106

First we note that −−→OX = −→OA + −−→AX. Next, you draw a parallelogram in the plane P whose sides are parallel to the vectors v and w , and whose diagonal is the line segment AX. The sides of this parallelogram represent vectors which are multiples of v and w and which add up to −−→AX. So, if one side of the parallelogram is sv and the other tw then we have −−→AX = sv + tw . With −−→OX = −→OA + −−→AX this implies (55).

44. Vector Bases

44.1. The Standard Basis Vectors. The notation for vectors which we have been using so far is not the most traditional. In the late 19th century Gibbs and Heavyside adapted Hamilton’s theory of Quaternions to deal with vectors. Their notation is still popular in texts on electromagnetism and ﬂuid mechanics. Deﬁne the following three vectors: 1 0 0 i = 0 , j = 1 , k = 0 .       0 0 1       Then every vector can be written as a linear combination of i, j and k, namely as follows:

a1 a2 = a1i + a2j + a3k.   a3   Moreover, there is only one way to write a given vector as a linear combination of i,j, k . { } This means that

a1 = b1

a1i + a2j + a3k = b1i + b2j + b3k a2 = b2 ⇐⇒   a3 = b3

For plane vectors one deﬁnes  1 0 i = , j = 0 1 and just as for three dimensional vectors one can write every (plane) vector a as a linear combination of i and j, a1 = a1i + a2j. a2 Just as for space vectors, there is only one way to write a given vector as a linear combination of i and j.

44.2. A Basis of Vectors (in general)*. The vectors i,j, k are called the standard basis vectors. They are an example of what is called a “basis”. Here is the deﬁnition in the case of space vectors:

44.3. Definition. A triple of space vectors u , v, w is a basis if every space vector { } a can be written as a linear combination of u , v, w , i.e. { } a = auu + avv + aww , and if there is only one way to do so for any given vector a (i.e. the vector a determines the coefficients au,av,aw). For plane vectors the definition of a basis is almost the same, except that a basis consists of two vectors rather than three: 107

44.4. Deﬁnition. A pair of plane vectors u , v is a basis if every plane vector a { } can be written as a linear combination of u , v , i.e. a = auu + avv, and if there is { } only one way to do so for any given vector a (i.e. the vector a determines the coeﬃcients au,av).

45. Dot Product

45.1. Deﬁnition. The “inner product” or “dot product” of two vectors is given by

a1 b1 a2 b2 = a1b1 + a2b2 + a3b3.     a3 b3     Note that the dot-product of two vectors is a number! The dot product of two plane vectors is (predictably) deﬁned by

a1 b1 = a1b1 + a2b2. a2 b2 An important property of the dot product is its relation with the length of a vector: (56) a 2 = a a.

45.2. Algebraic properties of the dot product. The dot product satisﬁes the following rules, (57) a b = b a (58) a (b + c)= a b + a c (59) (b + c) a = b a + c a (60) t(a b)=(ta) b which hold for all vectors a,b,c and any real number t.

45.3. Example. Simplify a + b 2. One has a + b 2 =(a + b) (a + b) = a (a + b)+ b (a + b) = a a + a b + b a +b b =2a b by (57) = a 2 + 2a b+ b 2

45.4. The diagonals of a parallelogram. Here is an example of how you can use the algebra of the dot product to prove something in geometry. Suppose you have a parallelogram one of whose vertices is the origin. Label the vertices, starting at the origin and going around counterclockwise, O, A, C and B. Let a = −→OA, b = −−→OB, c = −−→OC. One has

−−→OC = c = a + b, and −→AB = b a. − These vectors correspond to the diagonals OC and AB 108

45.5. Theorem. In a parallelogram OACB the sum of the squares of the lengths of the two diagonals equals the sum of the squares of the lengths of all four sides.

Proof. The squared lengths of the diagonals are −−→OC 2 = a + b 2 = a 2 + 2a b + b 2 −→AB 2 = a b 2 = a 2 2a b + b 2 − − Adding both these equations you get −−→OC 2 + −→AB 2 = 2 a 2 + b 2 . The squared lengths of the sides are −→OA 2 = a 2, −→AB 2 = b 2, −−→BC 2 = a 2, −−→OC 2 = b 2. Together these also add up to 2 a 2 + b 2 .

Figure 20. Proof of the law of cosines

45.6. The dot product and the angle between two vectors. Here is the most important interpretation of the dot product:

45.7. Theorem. If the angle between two vectors a and b is θ, then one has a b = a b cos θ.

Proof. We need the law of cosines from high-school trigonometry. Recall that for a triangle OAB with angle θ at the point O, and with sides OA and OB of lengths a and b, the length c of the opposing side AB is given by (61) c2 = a2 + b2 2ab cos θ. − In trigonometry this is proved by dropping a perpendicular line from B onto the side OA. The triangle OAB gets divided into two right triangles, one of which has AB as hypotenuse. Pythagoras then implies c2 =(b sin θ)2 +(a b cos θ)2 . − After simpliﬁcation you get (61). To prove the theorem you let O be the origin, and then observe that the length of the side AB is the length of the vector −→AB = b a. Here a = −→OA, b = −−→OB, and hence − c2 = b a 2 =(b a) (b a)= b 2 + a 2 2a b. − − − − 109

Compare this with (61), keeping in mind that a = a and b = b : you are led to conclude that 2a b = 2ab cos θ, and thus a b = a b cos θ. − −

45.8. Orthogonal projection of one vector onto an- λa = other. The following construction comes up very often. Let x// a = 0 be a given vector. Then for any other vector x there is a number λ such that x a y x = λ + x⊥ where y a. In other words, you can write any vector x as the a ⊥ sum of one vector parallel to a and another vector orthogonal Given x and a, ﬁnd x to a. The two vectors λa and y are called the parallel and ⊥ and x//. orthogonal components of the vector x (with respect to a), and sometimes the following notation is used

// x = λa, x⊥ = y, so that // x = x + x⊥. // There are moderately simple formulas for x and x⊥, but it is better to remember the following derivation of these formulas. Assume that the vectors a and x are given. Then we look for a number λ such that y = x λa is perpendicular to a. Recall that a (x λa) if and only if − ⊥ − a (x λa) = 0. − Expand the dot product and you get this equation for λ a x λa a = 0, − whence a x a x (62) λ = = a a a 2 To compute the parallel and orthogonal components of x w.r.t. a you first compute λ according to (62), which tells you that the parallel component is given by a x x// = λa = a. a a // The orthogonal component is then “the rest,” i.e. by definition x ⊥ = x x , so − // a x x⊥ = x x = x a. − − a a 45.9. Defining equations of lines. In 43 we saw how to generate points on a § line given two points on that line by means of a “parametrization.” I.e. given points A and B on the line ℓ the point whose position vector is x = a + t(b a) will be on ℓ for − any value of the “parameter” t. 110

In this section we will use the dot-product to give a X different description of lines in the plane (and planes in three dimensional space.) We will derive an equation for a line. n Rather than generating points on the line ℓ this equation tells us if any given point X in the plane is on the line or A x not. Here is the derivation of the equation of a line in the ℓ a plane. To produce the equation you need two ingredients: 1. One particular point on the line (let’s call this point O A, and write a for its position vector), Is X on ℓ? 2. a normal vector n for the line, i.e. a nonzero vector which is perpendicular to the line. Now let X be any point in the plane, and consider the line segment AX. Clearly, X will be on the line if and only if AX is parallel to ℓ 8 • Since ℓ is perpendicular to n , the segment AX and the line ℓ will be parallel if • and only if AX n . ⊥ AX n holds if and only if −−→AX n = 0. • ⊥ So in the end we see that X lies on the line ℓ if and only if the following vector equation is satisfied: (63) −−→AX n =0 or (x a) n = 0 − This equation is called a defining equation for the line ℓ. Any given line has many defining equations. Just by changing the length of the normal you get a different equation, which still describes the same line.

45.10. Line through one point and perpendicular to another. Find a deﬁning equation for the line ℓ which goes through A(1, 1) and is perpendicular to the line segment AB where B is the point (3, 1). − Solution. We already know a point on the line, namely ℓ A, but we still need a normal vector. The line is required to be perpendicular to AB, so n = −→AB is a normal vector: 2 3 1 2 n = −→AB = − = 1 A ( 1) 1 2 − − − 0 Of course any multiple of n is also a normal vector, for in- 1 2 3 stance −1 1 B m = 1 n = 2 1 −2 − is a normal vector. 1 With a =( 1 ) we then get the following equation for ℓ

2 x1 1 n (x a)= − = 2x1 2x2 = 0. − 2 x2 1 − − − If you choose the normal m instead, you get

1 x1 1 m (x a)= − = x1 x2 = 0. − 1 x2 1 − − − Both equations 2x1 2x2 = 0 and x1 x2 = 0 are equivalent. − − 8 From plane Euclidean geometry: parallel lines either don’t intersect or they coincide. 111

45.11. Distance to a line. Let ℓ be a line in the plane and assume a point A on the line as well as a vector n perpendicular to ℓ are known. Using the dot product one can easily compute the distance from the line to any other given point P in the plane. Here is how: Draw the line m through A perpendicular to ℓ, and drop a perpendicular line from P onto m. let Q be the projection of P onto m. The distance from P to ℓ is then equal to the length of the line segment AQ. Since AQP is a right triangle one has AQ = AP cos θ. Here θ is the angle between the normal n and the vector −→AP . One also has n (p a)= n −→AP = −→AP n cos θ = AP n cos θ. − Hence we get n (p a) dist(P,ℓ)= − . n P dist(P,ℓ) ℓ dist(P,ℓ) ℓ m θ P Q θ n A A n Q m π π θ < θ> 2 2 This argument from a drawing contains a hidden assumption, namely that the point P lies on the side of the line ℓ pointed to by the vector n . If this is not the case, so that n and −→AP point to opposite sides of ℓ, then the angle between them exceeds 90◦, i.e. θ > π/2. In this case cos θ < 0, and one has AQ = AP cos θ. the distance formula therefore has − to be modified to n (p a) dist(P,ℓ)= − . − n X 45.12. Defining equation of a plane. Just as n we have seen how you can form the defining equation −−→AX for a line in the plane from just one point on the line and one normal vector to the line, you can also form the A x defining equation for a plane in space, again knowing only one point on the plane, and a vector perpendicular a to it. If A is a point on some plane P and n is a vector perpendicular to P, then any other point X lies on P if and only if −−→AX n . In other ⊥ words, in terms of the position vectors a and x of A and X, the point X is on P n (x a) = 0. ⇐⇒ − Arguing just as in 45.11 you find that the distance § of a point X in space to the plane P is n (x a) (64) dist(X, P)= − . ± n 112

Here the sign is “+” if X and the normal n are on the same side of the plane P; otherwise the sign is “ ”. − 45.13. Example. Find the deﬁning equation for the plane P through the point 1 A(1, 0, 2) which is perpendicular to the vector 2 . 1 1 Solution: We know a point (A) and a normal vector n = 2 for P. Then any point 1 x1 X with coordinates (x1,x2,x3), or, with position vector x = x2 , will lie on the plane x3 P if and only if

1 x1 1 n (x a) = 0 2 x2 0 = 0 − ⇐⇒     −   1  x3 2        1 x1 1  −  2 x2 = 0 ⇐⇒     1 x3 2 − 1 (x1 1) + 2(x2) + 1 (x3 2) = 0 ⇐⇒ − − x1 + 2x2 + x3 3 = 0. ⇐⇒ −

x3 3

1 1 n = 2 1 1 2 3

x2 1

45.14. Example continued. Let P be the plane from the previous example. Which of the points P (0, 0, 1), Q(0, 0, 2), R( 1, 2, 0) and S( 1, 0, 5) lie on P? Compute the − − distances from the points P,Q,R,S to the plane P. Separate the points which do not lie on P into two group of points which lie on the same side of P. Solution: We apply (64) to the position vectors p, q, r,s of the points P,Q,R,S. For each calculation we need n = 12 + 22 + 12 = √6. 1 The third component of the given normal n = 2 is positive, so n points “upwards.” 1 Therefore, if a point lies on the side of P pointed to by n , we shall say that the point lies above the plane. 113

0 1 P : p = 0 , p a = −0 , n (p a) = 1 ( 1) + 2 (0) + 1 ( 1) = 2 1 − 1 − − − − − n (p a) 2 1 − = = √6. n − √6 − 3 P P 1 This quantity is negative, so P lies below . Its distance to is 3 √6. 0 1 Q: q = 0 , p a = −0 , n (p a) = 1 ( 1) + 2 (0) + 1 (0) = 1 2 − 0 − − − n (p a) 1 1 − = = √6. n − √6 − 6 P P 1 This quantity is negative, so Q also lies below . Its distance to is 6 √6. 1 2 R: r = −2 , p a = −2 , n (p a) = 1 ( 2) + 2 (2) + 1 ( 2) = 0 0 − 2 − − − − n (p a) − = 0. n Thus R lies on the plane P, and its distance to P is of course 0. 1 2 S: s = −0 , p a = −0 , n (p a) = 1 ( 1) + 2 (0) + 1 (3) = 2 5 − 3 − − n (p a) 2 1 − = = √6. n √6 3 P P 1 This quantity is positive, so S lies above . Its distance to is 3 √6. We have found that P and Q lie below the plane, R lies on the plane, and S is above the plane.

45.15. Where does the line through the points B(2, 0, 0) and C(0, 1, 2) intersect the plane P from example 45.13? Solution: Let ℓ be the line through B and C. We set up the parametric equation for ℓ. According to 43, (54) every point X on ℓ § has position vector x given by 2 0 2 2 2t − − (65) x = b + t(c b)= 0 + t 1 0 = t −    −    0 2 0 2t −       for some value of t. The point X whose position vector x is given above lies on the plane P if x satisfies the defining equation of the plane. In example 45.13 we found this defining equation. It was

(66) n (x a) = 0, i.e. x1 + 2x2 + x3 3 = 0. − − So to ﬁnd the point of intersection of ℓ and P you substitute the parametrization (65) in the deﬁning equation (66):

0= x1 + 2x2 + x3 3=(2 2t)+2(t)+(2t) 3 = 2t 1. − − − − 1 This implies t = 2 , and thus the intersection point has position vector 2 2t 1 − x = b + 1 (c b)= t = 1 , 2 −    2  2t 1     P 1 i.e. ℓ and intersect at X(1, 2 , 1). 114

46. Cross Product

46.1. Algebraic definition of the cross product. Here is the definition of the cross-product of two vectors. The definition looks a bit strange and arbitrary at first sight – it really makes you wonder who thought of this. We will just put up with that for now and explore the properties of the cross product. Later on we will see a geometric interpretation of the cross product which will show that this particular definition is really useful. We will also find a few tricks that will help you reproduce the formula without memorizing it.

46.2. Deﬁnition. The “outer product” or “cross product” of two vectors is given by

a1 b1 a2b3 a3b2 − a2 b2 = a3b1 a1b3   ×    −  a3 b3 a1b2 a2b1 − Note that the cross-product of two vectors  is again a vector!

46.3. Example. If you set b = a in the definition you find the following important fact: The cross product of any vector with itself is the zero vector: a a = 0 for any vector a. × 1 2 46.4. Example. Let a = 2 , b = −1 and compute the cross product of these 3 0 vectors. Solution: 1 2 2 0 3 1 3 − − − i j k a b = 2 1 = 3 ( 2) 1 0 = 6 × ×   ×    − −  −  3 0 1 1 2 ( 2) 5 i 0 k j  − −  −         j k 0 i In terms of the standard basis vectors you can check the multiplication table. An easy − k j i 0 way to remember the multiplication table is to put the vectors i,j, k clockwise in a circle. − Given two of the three vectors their product is either plus or minus the remaining vector. i To determine the sign you step from the first vector to the second, to the third: if this k j makes you go clockwise you have a plus sign, if you have to go counterclockwise, you get a minus. The products of i,j and k are all you need to know to compute the cross product. Given two vectors a and b write them as a = a1i + a2j + a3k and b = b1i + b2j + b3k, and multiply as follows

a b =(a1i + a2j + a3k) (b1i + b2j + b3k) × × = a1i (b1i + b2j + b3k) × +a2j (b1i + b2j + b3k) × +a3k (b1i + b2j + b3k) × = a1b1i i + a1b2i j + a1b3i k + × × × a2b1j i + a2b2j j + a2b3j k + × × × a3b1k i + a3b2k j + a3b3k k × × × = a1b10 + a1b2k a1b3j − a2b1k + a2b20 + a2b3i + − a3b1j a3b2i + a3b30 − =(a2b3 a3b2)i +(a3b1 a1b3)j +(a1b2 a2b1)k − − − 115

This is a useful way of remembering how to compute the cross product, particularly when many of the components ai and bj are zero.

46.5. Example. Compute k (pi + qj + rk): × k (pi + qj + rk)= p(k i)+ q(k j)+ r(k k)= qi + pj. × × × × − There is another way of remembering how to find a b. It involves the “triple × product” and determinants. See 46.7. § 46.6. Algebraic properties of the cross product. Unlike the dot product, the cross product of two vectors behaves much less like ordinary multiplication. To begin with, the product is not commutative – instead one has (67) a b = b a for all vectors a and b. × − × This property is sometimes called “anti-commutative.” Since the crossproduct of two vectors is again i (i j)= i k = j, a vector you can compute the cross product of × × × − (i i) j = 0 j = 0 three vectors a, b,c. You now have a choice: do × × × so “ ” is not associative you first multiply a and b, or b and c, or a and c? × With numbers it makes no difference (e.g. 2 (3 5) = 2 15 = 30 and (2 3) 5 = 6 5= × × × × × × also 30) but with the cross product of vectors it does matter: the cross product is not associative, i.e. a (b c) ==(a b) c for most vectors a,b,c. × × × × The distributive law does hold, i.e. a (b + c)= a b + a c, and (b + c) a = b a + c a × × × × × × is true for all vectors a,b,c. Also, an associative law, where one of the factors is a number and the other two are vectors, does hold. I.e. t(a b)=(ta) b = a (tb) × × × holds for all vectors a,b and any number t. We were already using these properties when we multiplied (a1i + a2j + a3k) (b1i + b2j + b3k) in the previous section. × Finally, the cross product is only defined for space vectors, not for plane vectors.

46.7. The triple product and determinants.

46.8. Deﬁnition. The triple product of three given vectors a,b, and c is deﬁned to be a (b c). × In terms of the components of a,b, and c one has

a1 b2c3 b3c2 − a (b c)= a2 b3c1 b1c3 ×    −  a3 b1c2 b2c1 − = a1b2c3 a1b3c2 + a2b3c1 a2b1c3 + a3b1c2 a3b2c1. − − − This quantity is called a determinant, and is written as follows

a1 b1 c1 (68) a2 b2 c2 = a1b2c3 a1b3c2 + a2b3c1 a2b1c3 + a3b1c2 a3b2c1 − − − a3 b3 c3 116

There’s a useful shortcut for computing such a de- a1 b1 c1 a1 b1 terminant: after writing the determinant, append a fourth and a ﬁfth column which are just copies of the a2 b2 c2 a2 b2 ﬁrst two columns of the determinant. The determinant a b c a b then is the sum of six products, one for each dotted line 3 3 3 3 3 in the drawing. Each term has a sign: if the factors are − − − + + + read from top-left to bottom-right, the term is positive, if they are read from top-right to bottom left the term is negative. This shortcut is also very useful for computing the crossproduct. To compute the cross product of two given vectors a and b you arrange their components in the following determinant

i a1 b1 (69) a b = j a2 b2 =(a2b3 a3b2)i +(a3b1 a1b3)j +(a1b2 a2b1)k. × − − − k a3 b3 This is not a normal determinant since some of its entries are vectors, but if you ignore that odd circumstance and simply compute the determinant according to the deﬁnition (68), you get (69). An important property of the triple product is that it is much more symmetric in the factors a,b,c than the notation a (b c) suggests. ×

46.9. Theorem. For any triple of vectors a,b,c one has

a (b c)= b (c a)= c (a b), × × × and a (b c)= b (a c)= c (b a). × − × − × In other words, if you exchange two factors in the product a (b c) it changes its sign. If × you “rotate the factors,” i.e. if you replace a by b, b by c and c by a, the product doesn’t change at all.

46.10. Geometric description of the cross product.

46.11. Theorem. a b a,b × ⊥

a b Proof. We use the triple product: × b θ a (a b)= b (a a)= 0 × × a since a a = 0 for any vector a. It follows that a b is × × perpendicular to a.

Similarly, b (a b) = a (b b) = 0 shows that a b is × × perpendicular to b. 117

46.12. Theorem. a b = a b sin θ ×

Proof. Bruce9 just slipped us a piece of paper with the following formula on it: (70) a b 2 +(a b)2 = a 2 b 2. × a1 b1 After setting a = a2 and b = b2 and diligently computing both sides we find that a3 b 3 this formula actually holds for any pair of vectors a,b! The (long) computation which implies this identity will be presented in class (maybe). If we assume that Lagrange’s identity holds then we get a b 2 = a 2 b 2 (a b)2 = a 2 b 2 a 2 b 2 cos2 θ = a 2 b 2 sin2 θ × − − since 1 cos2 θ = sin2 θ. The theorem is proved. − These two theorems almost allow you to construct the cross product of two vectors geometrically. If a and b are two vectors, then their cross product satisfies the following description: (1) If a and b are parallel, then the angle θ between them vanishes, and so their cross product is the zero vector. Assume from here on that a and b are not parallel. (2) a b is perpendicular to both a and b. In other words, since a and b are not par- × allel, they determine a plane, and their cross product is a vector perpendicular to this plane. (3) the length of the cross product a b is a b sin θ. × There are only two vectors that satisfy conditions 2 and a b × 3: to determine which one of these is the cross product you must apply the Right Hand Rule (screwdriver rule, corkscrew b rule, etc.) for a,b, a b: if you turn a screw whose axis is × a perpendicular to a and b in the direction from a to b, the screw moves in the direction of a b. × Alternatively, without seriously injuring yourself, you should be able to make a fist with your right hand, and then stick out your thumb, index and middle fingers so that your thumb is a, your index finger is b and your middle finger is a b. Only people with the most flexible × joints can do this with their left hand.

47. A few applications of the cross product

47.1. Area of a parallelogram. Let ABCD be a parallelogram. Its area is given by “height times base,” a formula which should be familiar from high school geometry. If the angle between the sides AB and AD is θ, then the height of the parallelogram is −−→AD sin θ, so that the area of ABCD is (71) area of ABCD = −→AB −−→AD sin θ = −→AB −−→AD . × The area of the triangle ABD is of course half as much, area of triangle ABD = 1 −→AB −−→AD . 2 × 9It’s actually called Lagrange’s identity. Yes, the same Lagrange who found the formula for the remainder term. 118

These formulae are valid even when the points A,B,C, and D are points in space. Of course they must lie in one plane for otherwise ABCD couldn’t be a parallelogram.

47.2. Example. Let the points A(1, 0, 2), B(2, 0, 0), C(3, 1, 1) and D(2, 1, 1) be − given. Show that ABCD is a parallelogram, and compute its area.

Solution: ABCD will be a parallelogram if and only if −→AC = −→AB + −−→AD. In terms of the position vectors a,b, c and d of A,B,C,D this boils down to

c a =(b a)+(d a), i.e. a + c = b + d. − − − For our points we get

1 3 4 2 2 4 a + c = 0 + 1 = 1 , b + d = 0 + 1 = 1 .             2 1 1 0 1 1 −             So ABCD is indeed a parallelogram. Its area is the length of

2 1 2 1 1 1 2 − − − −→AB −−→AD = 0 1 0 = 0 1 = 1 . ×   ×  −    × −  −  × 0 2 × 1 2 2 × 1 1 − − − − −           So the area of ABCD is ( 2)2 +( 1)2 +( 1)2 = √6. − − − n = a b × b a 47.3. Finding the normal to a plane. If you know two vectors a and b which are parallel to a given plane P but not parallel to each other, then you can find a normal vector for the plane P by computing n = a b. × We have just seen that the vector n must be perpendicular to both a and b, and hence10 it is perpendicular to the plane P. This trick is especially useful when you have three points A, B and C, and you want to find the defining equation for the plane P through these points. We will assume that the three points do not all lie on one line, for otherwise there are many planes through A, B and C. To find the defining equation we need one point on the plane (we have three of them), and a normal vector to the plane. A normal vector can be obtained by computing the cross product of two vectors parallel to the plane. Since −→AB and −→AC are both parallel to P, the vector n = −→AB −→AC is such a normal vector. × Thus the defining equation for the plane through three given points A, B and C is

n (x a) = 0, with n = −→AB −→AC =(b a) (c a). − × − × −

10This statement needs a proof which we will skip. Instead have a look at the picture 119

47.4. Example. Find the defining equation of the plane P through the points A(2, 1, 0), − B(2, 1, 1) and C( 1, 1, 1). Find the intersections of P with the three coordinate axes, − − and find the distance from the origin to P. Solution: We have 0 3 − −→AB = 2 and −→AC = 2     1 1 − so that     0 3 4 − n = −→AB −→AC = 2 2 = 3 ×   ×     × 1 × 1 6 − is a normal to the plane. The defining equation  forP is therefore 

4 x1 2 − 0= n (x a)= 3 x2 + 1 −     6 x3 0 − i.e.     4x1 + 3x2 + 6x3 5 = 0. − The plane intersects the x1 axis when x2 = x3 = 0 and hence 4x1 5 = 0, i.e. in the point 5 5 − 5 ( 4 , 0, 0). The intersections with the other two axes are (0, 3 , 0) and (0, 0, 6 ). The distance from any point with position vector x to P is given by n (x a) dist = − , ± n 0 so the distance from the origin (whose position vector is x = 0 = 0 ) to P is 0 a n 2 4+( 1) 3 + 0 6 5 distance origin to P = = − = ( 1.024 ). ± n ± √42 + 32 + 62 √61 ≈ 47.5. Volume of a parallelepiped. G H E F H C D G E B A D height F C base A B

A parallelepiped is a three dimensional body whose sides are parallelograms. For instance, a cube is an example of a parallelepiped; a rectangular block (whose faces are rectangles, meeting at right angles) is also a parallelepiped. Any parallelepiped has 8 vertices (corner points), 12 edges and 6 faces. ABCD Let EFGH be a parallelepiped. If we call one of the faces, say ABCD, the base of the parallelepiped, then the other face EFGH is parallel to the base. The height of the parallelepiped is the distance from any point in EFGH to the base, e.g. to compute the ABCD height of EFGH one could compute the distance from the point E (or F , or G, or H) to the plane through ABCD. ABCD The volume of the parallelepiped EFGH is given by the formula ABCD Volume = Area of base height. EFGH × 120

Since the base is a parallelogram we know its area is given by Area of baseABCD = −→AB −−→AD × We also know that n = −→AB −−→AD is a vector perpendicular to the plane through ABCD, × i.e. perpendicular to the base of the parallelepiped. If we let the angle between the edge AE and the normal n be ψ, then the height of the parallelepiped is given by height = −→AE cos ψ. Therefore the triple product of −→AB, −−→AD, −→AE is ABCD Volume = height Area of base EFGH × = −→AE cos ψ −→AB −−→AD , × i.e. ABCD Volume = −→AE (−→AB −−→AD). EFGH ×

48. Notation

In the next chapter we will be using vectors, so let’s take a minute to summarize the concepts and notation we have been using. Given a point in the plane, or in space you can form its position vector. So associated to a point we have three diﬀerent objects: the point, its position vector and its coordinates. here is the notation we use for these:

Object Notation

Point...... Upper case letters, A, B, etc. Position vector ...... Lowercase letters with an arrow on top. The position vector −→OA of the point A should be a, so that letters match across changes from upper to lower case. Coordinates of a point. . . The coordinates of the point A are the same as the components of its position vector a: we use lower case letters with a subscript to indicate which coordinate we have in mind: (a1,a2).

49. PROBLEMS

COMPUTING AND DRAWING VECTORS

426. Simplify the following 427. If a,b,c are as in the previous prob- 1 0 lem, then which of the following expres- a = 2 + 3 1 ; sions mean anything? Compute those −    3 3 expressions that are well deﬁned.  1  4 b = 12 3 ; 1/3 − 1 1 1 i ii iii c = (1+ t) t , ( ) a + b ( ) b + c ( ) πa 1 t − t 2 − − (iv) b (v) b/c (vi) a + b 1 0 0 2 d = t 0 + t2 1 0 . (vii) b (viii) b/ c   −  −   0 2 1       121

1 2 432. Given points A(2, 1) and B( 1, 4) 428. Let a = 2 and b = 1 . −  −   −  compute the vector −→AB. Is −→AB a po- 2 1 sition vector? Compute:     433. Given: points A(2, 1), B(3, 2), C(4, 4) (1) a || || and D(5, 2). Is ABCD a parallelo- (2) 2a gram? (3) 2a 2 || || (4) a +b 434. Given: points A(0, 2, 1), B(0, 3, 2), (5) 3a b C(4, 1, 4) and D. − (a) If ABCD is a parallelogram, then 429. Let u , v, w be three given vectors, what are the coordinates of the point D? and suppose (b) If ABDC is a parallelogram, then a = v+w , b = 2u w , c = u +v+w . − what are the coordinates of the point (a) Simplify p = a + 3b c and q = D? − c 2(u + a). − 435. You are given three points in the (b) Find numbers r,s,t such that ra + plane: A has coordinates (2, 3), B has sb + tc = u . coordinates ( 1, 2) and C has coordi- − nates (4, 1). (c) Find numbers k,l,m such that ka + − lb + mc = v. (a) Compute the vectors −→AB, −→BA, −→AC, 430. Prove the Algebraic Properties (50), −→CA, −−→BC and −−→CB. (51), (52), and (53) in section 42.5. (b) Find the points P,Q,R and S whose 431. (a) Does there exist a number x such position vectors are −→AB, −→BA, −→AC, and that −−→BC, respectively. Make a precise draw- 1 x 2 + = ? ing in ﬁgure 21. 2 x 1 436. Have a look at ﬁgure 22 (b) Make a drawing of all points P whose (a) Draw the vectors 2v + 1 w , 1 v +w , position vectors are given by 2 − 2 and 3 v 1 w 1 x 2 − 2 p = + . 2 x (b) Find real numbers s,t such that sv + tw = a. (c) Do there exist a numbers x and y (c) Find real numbers p, q such that such that pv + qw = b. 1 1 2 x + y = ? 2 1 1 (d) Find real numbers k,l,m,n such that v = ka + lb, and w = ma + nw .

PARAMETRIC EQUATIONS FOR A LINE 122

Figure 21. Your drawing for problem 435

437. In the ﬁgure above draw the points 439. (a) Find a parametric equation for the whose position vectors are given by x = line ℓ through the points A(3, 0, 1) and a + t(b a) for t = 0, 1, 1 , 3 , 1, 2. (as B(2, 1, 2). − 3 4 − always, a = −→OA, etc.) (b) Where does ℓ intersect the coordinate planes? 438. In the ﬁgure above also draw the points whose position vector are given by 440. (a) Find a parametric equation for the x = b + s(a b) for s = 0, 1, 1 , 3 , 1, 2. − 3 4 − line which contains the two vectors

w a

Figure 22. Drawing for problem 436 123

2 3 442. Group problem. Let ABCD be a a = 3 and b = 2 .     tetrahedron, and let a, b,c, d be the po- 1 3 sition vectors of the points A,B,C,D.     i c1 ( ) Find position vectors of the midpoint (b) The vector c = 1 is on P of AB, the midpoint Q of CD and the   c3 midpoint M of PQ. this line. What is c?   (ii) Find position vectors of the midpoint R of BC, the midpoint S of AD 441. Group problem. Consider a trian- and the midpoint N of RS. gle ABC and let a,b,c be the position vectors of A, B, and C. D (a) Compute the position vector of the midpoint P of the line segment BC. Also compute the position vectors of the midpoints Q of AC and R of AB. (Make C a drawing.) (b) Let M be the point on the line segment AP which is twice as far from A as it is from P . Find the position vector of M.

ORTHOGONAL DECOMPOSITION OF ONE VECTOR WITH RESPECT TO ANOTHER

2 443. Given the vectors a = 1 and 3 1 // // b = 1 ﬁnd a , a⊥, b , b⊥ for which 0 // // a = a + a⊥, with a //b,a⊥ b, ⊥ and

// // b = b + b⊥, with b //a,b⊥ a. ⊥ 445. An eraser is lying on the plane P with equation x1 +3x2 +x3 = 6. Gravity pulls the eraser down, and exerts a force given 444. Bruce left his backpack on a hill, by which in some coordinate system hap- 0 pens to be the line with equation 12x + f = 0 . 1 grav   5x = 130. mg 2 −   P The force exerted by gravity on the (a) Find a normal n for the plane . 0 backpack is f grav = mg . Decompose (b) Decompose the force f into a this force into a part− perpendicular to part perpendicular to the plane P and a the hill, and a part parallel to the hill. part perpendicular to n .

THE DOT PRODUCT 124

446. (i) Simplify a b 2. 451. Let ℓ and m be the lines with − (ii) Simplify 2a b 2. parametrizations − (iii) If a has length 3, b has length 2 1 0 2 7 and a b = 2, then compute a + b , ℓ : x = +t , m : x = +s − − 0 2 1 3 a b and 2a b . − − − 447. Simplify (a + b) (a b). Where do they intersect, and ﬁnd the − angle between ℓ and m. 448. Find the lengths of the sides, and the angles in the triangle ABC whose ver- 452. Let ℓ and m be the lines with tices are A(2, 1), B(3, 2), and C(1, 4). parametrizations 449. Group problem. Given: A(1, 1), B(3, 2) and a point C which lies on 2 1 0 2 − the line with parametric equation c = ℓ : x = 1 +t 2 , m : x = 1 +s 0 0 1         ( 3 )+ t 1 . If ABC is a right trian- 4 0 1 3 − △ gle, then where is C? (There are three − −         possible answers, depending on whether Do ℓ and m intersect? Find the angle you assume A, B or C is the right an- between ℓ and m. gle.) 450. (i) Find the deﬁning equation and a 453. Let ℓ and m be the lines with normal vector n for the line ℓ which is parametrizations 1 the graph of y =1+ 2 x. ii 2 1 0 2 ( ) What is the distance from the origin − ℓ : x = α +t 2 , m : x = 1 +s 0 to ℓ?         1 0 1 3 (iii) Answer the same two questions for −         the line m which is the graph of y = Here α is some unknown number. 2 3x. − (iv) What is the angle between ℓ and If it is known that the lines ℓ and m m? intersect, what can you say about α?

THE CROSS PRODUCT

454. Compute the following cross products 455. Compute the following cross products

(i) i (i + j) 3 3 × (i) 1 2   ×   (ii) (√2i + j) √2j 2 1 ×     (iii) (2i + k) (j k) × − 12 12 (ii) 71 71 (iv) (cos θi + sin θk) (sin θi cos θk) − 1  × − 1  × − 3 2 3 2     456. (i) Simplify (a + b) (a + b). × 1 1 (iii) 0 1 (ii) Simplify (a b) (a b).   ×   − × − 0 0     (iii) Simplify (a + b) (a b). × − √2 0 (iv) 1 √2 457. True or False: If a b = c b and   ×   × × 0 0 b = 0 then a = c?     125

458. Group problem. Given A(2, 0, 0), B(0, 0, 2) and C(2, 2, 2). Let P be the E H plane through A, B and C. (i) Find a normal vector for P.

(ii) Find a defining equation for P. F G (iii) What is the distance from D(0, 2, 0) D A to P? What is the distance from the origin O(0, 0, 0) to P? B C (iv) Do D and O lie on the same side of P? (v) Find the area of the triangle ABC. (vi) Where does the plane P intersect ABCD 462. Group problem. Let EFGH be the the three coordinate axes? cube with A at the origin, B(1, 0, 0), 459. (i) Does D(2, 1, 3) lie on the plane P D(0, 1, 0) and E(0, 0, 1). through the points A( 1, 0, 0), B(0, 2, 1) − and C(0, 3, 0)? (i) Find the coordinates of all the points (ii) The point E(1, 1, α) lies on P. What A, B, C, D, E, F , G, H. is α? ii 460. Given points A(1, 1, 1), B(2, 0, 1) ( ) Find the position vectors of the mid- − and C(1, 2, 0). points of the line segments AG, BH, CE and DF . Make a drawing of the cube (i) Where is the point D which makes with these line segments. ABCD into a parallelogram? (ii) What is the area of the parallelo- (iii) Find the defining equation for the gram ABCD? plane BDE. Do the same for the plane iii ( ) Find a defining equation for the CFH. Show that these planes are par- P plane containing the parallelogram allel. ABCD. iv P ( ) Where does intersect the coordi- (iv) Find the parametric equation for nate axes? the line through AG. 461. Given points A(1, 0, 0), B(0, 2, 0) and D( 1, 0, 1) and E(0, 0, 2). v − ( ) Where do the planes BDE and i ABCD ( ) If P = EFGH is a parallelepiped, CFH intersect the line AG? then where are the points C,F,G and H? (vi) Find the angle between the planes (ii) Find the area of the base ABCD of BDE and BGH. P. iii ( ) Find the height of P. (vii) Find the angle between the planes (iv) Find the volume of P. BDE and BCH. Draw these planes. 126

Chapter 6: Vector Functions and Parametrized Curves

50. Parametric Curves

50.1. Deﬁnition. A vector function f of one variable is a function of one real variable, whose values f (t) are vectors. In other words for any value of t (from a domain of allowed values, usually an interval) the vector function f produces a vector f (t). Write f in components: f (t) f (t)= 1 . f2(t) The components of a vector function f of t are themselves functions of t. They are ordinary ﬁrst-semester-calculus-style functions. An example of a vector function is

2 2 t 2t 1 2(1) 1 f(t)= − 2 , so f(1) = − 2 = − 1 + cos πt 1 + (cos π) 2 (just to mention one.)

50.2. Deﬁnition. A parametric curve is a vector function x = x(t) of one real variable t. The variable t is called the parameter. Synonyms: “Parametrized curve,” or “parametrization,” or “vector function (of one variable).” Logically speaking a parametrized curve is the same thing as a vector function. The name “parametrized curve” is used to remind you of a very natural and common interpretation of the concept “parametric curve.” In this interpretation a vector function, or parametric curve x (t) describes the motion of a point in the plane or space. Here t stands for time, and x(t) is the position vector at time t of the moving point.

1.5

1.0 −1.0 3.5 2.0 −0.5 2.5 t=−1.5 0.5 3.0 t=0.0 A picture of a vector function.

Instead of writing a parametrized curve as a vector function, one sometimes speciﬁes the two (or three) components of the curve. Thus one will say that a parametric curve is given by

x1 = x1(t), x2 = x2(t), (and x3 = x3(t) if we have a space curve). 127

51. Examples of parametrized curves

51.1. An example of Rectilinear Motion. Here’s a parametric curve: 1+ t (72) x(t)= . 2 + 3t The components of this vector function are

(73) x1(t)=1+ t, x2(t)=2+3t. Both components are linear functions of time (i.e. the parameter t), so every time t increases by an amount ∆t (every time ∆t seconds go by) the ﬁrst component increases by ∆t, and the x2 component increases by 3∆t. So the point at x(t) moves horizontally to the left with speed 1, and it moves vertically upwards with speed 3.

Which curve is traced out by this vector function? In this example we can find out by eliminating the parameter, i.e. solve t = 1 one of the two equations (73) for t, and substitute the value of t you find in the other equation. Here you can solve x1 =1+ t for t, with result t = x1 1. From there you find that − t = 0 3∆t x2 =2+3t =2+3(x1 1) = 3x1 1. − − So for any t the vector x (t) is the position vector of a point on the line x2 = 3x1 1 (or, if you prefer the old fashioned x,y coordinates, ∆t − y = 3x 1). − Conclusion: This particular parametric curve traces out a straight line with equation x2 = 3x1 1, going from left to right. − 51.2. Rectilinear Motion in general. This example gener- t = 1 alizes the previous example. The parametric equation for a straight − line from the previous chapter x(t)= a + tv, is a parametric curve. We had v = b a in 43. At time t = 0 the object is at the point − § with position vector a, and every second (unit of time) the object translates by v. The vector v is the velocity vector of this motion. 1 1 In the first example we had a =( 2 ), and v =( 3 ).

51.3. Going back and forth on a straight line. Consider x(t)= a + sin(t)v. At each moment in time the object whose motion is described by this parametric curve ﬁnds itself on the straight line ℓ with parametric equation x = a+s(b a), where b = a+v. − However, instead of moving along the line from one end to the other, the point at x(t) keeps moving back and forth along ℓ between a + v and a v. − 51.4. Motion along a graph. Let y = f(x) be some function of one variable (deﬁned for x in some interval) and consider the parametric curve given by t x(t)= = ti + f(t)j. f(t) At any moment in time the point at x (t) has x1 coordinate equal to t, and x2 = f(t) = f(x1), since x1 = t. So this parametric curve describes motion on the graph of y = f(x) in which the horizontal coordinate increases at a constant rate. 128

x(θ) 51.5. The standard parametrization of a circle. Con- sider the parametric curve θ cos θ x (θ)= . sin θ The two components of this parametrization are

x1(θ) = cos θ, x2(θ)= sinθ, and they satisfy 2 2 2 2 x1(θ) + x2(θ) = cos θ + sin θ = 1, so that x (θ) always points at a point on the unit circle. As θ increases from to + the point will rotate through the circle, going around −∞ ∞ inﬁnitely often. Note that the point runs through the circle in the counterclockwise direction, which is the mathematician’s favorite way of running around in circles.

51.6. The Cycloid. The Free Ferris Wheel Foundation is an organization whose goal is to empower fairground ferris wheels to roam freely and thus realize their potential. With blatant disregard for the public, members of the F2WF will clandestinely unhinge ferris wheels, thereby setting them free to roll throughout the fairground and surroundings. Suppose we were to step into the bottom of a ferris wheel at the moment of its liberation: what would happen? Where would the wheel carry us? Let our position be the point X, and let its position vector at time t be x(t). The parametric curve x(t) which describes our motion is called the cycloid.

In this example we are given a description of a motion, but no formula for the parametrization x(t). We will have to derive this formula ourselves. The key to ﬁnd- ing x(t) is the fact that the arc AX on the wheel is exactly as long as the line segment OA on the ground (i.e. the x1 axis). The length of the arc AX is exactly the angle θ (“arc = radius times angle in radians”), so the x1 coordinate of A and hence the center C of the circle is θ. To ﬁnd X consider the right triangle BCX. Its hypothenuse is the radius of the circle, i.e. CX has length 1. The angle at C is θ, and therefore you get BX = sin θ, BC = cos θ, and x1 = OA BX = θ sin θ, x2 = AC BC = 1 cos θ. − − − − 129

So the parametric curve deﬁned in the beginning of this example is θ sin θ x(θ)= − . 1 cos θ − Here the angle θ is the parameter, and we can let it run from θ = to θ = . −∞ ∞

x3 θ = 4π

θ = 2π

x2 height = aθ θ

x1 θ = 0

51.7. A three dimensional example: the Helix. Consider the vector function cos θ x(θ)= sin θ   aθ where a> 0 is some constant.  

If you ignore the x3 component of this vector function you get the parametrization of the circle from example 51.5. So as the parameter θ runs from to + , the x1,x2 part −∞ ∞ of x(θ) runs around on the unit circle inﬁnitely often. While this happens the vertical component, i.e. x3(θ) increases steadily from to at a rate of a units per second. −∞ ∞

52. The derivative of a vector function

If x(t) is a vector function, then we define its derivative to be dx x(t + h) x(t) x′(t)= = lim − . h 0 dt → h This definition looks very much like the first-semester-calculus-definition of the derivative of a function, but for it to make sense in the context of vector functions we have to explain what the limit of a vector function is. By definition, for a vector function f (t)= f1(t) one has f2(t) f1(t) limt a f1(t) lim f (t) = lim = → t a t a f2(t) limt a f2(t) → → → In other words, to compute the limit of a vector function you just compute the limits of its components (that will be our definition.) 130

Let’s look at the deﬁnition of the velocity vector again. Since x(t + h) x(t) 1 x (t + h) x (t) − = 1 1 h h x2(t + h) − x2(t) x1(t + h) x1(t) − = h  x2(t + h) x2(t)  −  h  we have   x(t + h) x(t) x′(t) = lim − h 0 → h x1(t + h) x1(t) lim − h 0 = → h  x2(t + h) x2(t) lim − h 0 h   →  x′ (t) = 1 x′ (t) 2 So: To compute the derivative of a vector function you must diﬀerentiate its components.

52.1. Example. Compute the derivative of cos t t sin t x (t)= and of y(t)= − . sin t 1 cos t − Solution: d cos t sin t x′(t)= = − dt sin t cos t d t sin t 1 cos t y′(t)= − = − . dt 1 cos t sin t − 53. Higher derivatives and product rules

If you differentiate a vector function x(t) you get another vector function, namely x′(t), and you can try to differentiate that vector function again. If you succeed, the result is called the second derivative of x(t). All this is very similar to how the second (and higher) derivative of ordinary functions were defined in 1st semester calculus. One even uses the same notation:11 2 dx′(t) d x x1′′(t) x ′′(t)= = = . dt dt2 x′′(t) 2 53.1. Example. Compute the second derivative of cos t t sin t x (t)= and of y(t)= − . sin t 1 cos t − Solution: In example 52.1 we already found the first derivatives, so you can use those. You find d sin t cos t x′′(t)= − = − dt cos t sin t − d 1 cos t sin t y′′(t)= − = . dt sin t cos t Note that our standard parametrization x (t) of the circle satisfies

x′′(t)= x(t). − 11 Not every function has a derivative, so it may happen that you can ﬁnd x ′(t) but not x ′′(t) 131

After defining the derivative in first semester calculus one quickly introduces the various rules (sum, product, quotient, chain rules) which make it possible to compute derivatives without ever actually having to use the limit-of-difference-quotient-definition. For vector functions there are similar rules which also turn out to be useful. The Sum Rule holds. It says that if x (t) and y(t) are differentiable12 vector functions, then so is z(t)= x(t) y(t), and one has ± dx(t) y(t) dx(t) dy(t) ± = . dt dt ± dt The Product Rule also holds, but it is more complicated, because there are several different forms of multiplication when you have vector functions. The following three versions all hold: If x (t) and y(t) are differentiable vector functions and f(t) is an ordinary differentiable function, then df(t)x(t) dx(t) df(t) = f(t) + x(t) dt dt dt dx(t) y(t) dy(t) dx(t) = x (t) + y(t) dt dt dt dx(t) y(t) dy(t) dx(t) × = x (t) + y(t) dt × dt dt × I hope these formulae look plausible because they look like the old fashioned product rule, but even if they do, you still have to prove them before you can accept their validity. I will prove one of these in lecture. You will do some more as an exercise. As an example of how these properties get used, consider this theorem:

53.2. Theorem. Let f (t) be a vector function of constant length (i.e. f (t) is constant.) Then f ′(t) f (t). ⊥ Proof. If f is constant, then so is f (t) f (t)= f (t) 2. the derivative of a constant function is zero, so d d df (t) 0= f (t) 2 = f (t) f (t) = 2f (t) . dt dt dt So we see that f f ′ = 0 which means that f ′ f . ⊥

54. Interpretation of x ′(t) as the velocity vector

x(t + h) x(t) −

x(t + h) x(t)

Figure 23. The vector velocity of a motion in the plane.

12A vector function is diﬀerentiable if its derivative actually exists, i.e. if all its components are diﬀerentiable. 132

Let x(t) be some vector function and interpret it as describing the motion of some point in the plane (or space). At time t the point has position vector x(t); a little later, more precisely, h seconds later the point has position vector x(t + h). Its displacement is the diﬀerence vector x (t + h) x (t). − Its average velocity vector between times t and t + h is displacement vector x (t + h) x(t) = − . time lapse h

If the average velocity between times t and t + h converges to one deﬁnite vector as h 0, then this limit is a reasonable candidate for the velocity vector at time t of the → parametric curve x (t). Being a vector, the velocity vector has both magnitude and direction. The length of the velocity vector is called the speed of the parametric curve. We use the following notation: we always write v(t)= x′(t) for the velocity vector, and v(t)= v(t) = x ′(t) for its length, i.e. the speed. The speed v is always a nonnegative number; the velocity is always a vector.

54.1. Velocity of linear motion. If x (t)= a + tv, as in examples 51.1 and 51.2, then a + tv x(t)= 1 1 a + tv 2 2 so that v1 x′(t)= = v. v 2 So when you represent a line by a parametric equation x (t)= a + tv, the vector v is the velocity vector. The length of v is the speed of the motion. 1 1+t In example 51.1 we had v =( 3 ), so the speed with which the point at x(t)= 1+3t traces out the line is v = v = √12 + 32 = √10. x(t) 54.2. Motion on a circle. Consider the parametrization ωt R cos ωt v(t) R x(t)= . R sin ωt The point X at x (t) is on the circle centered at the origin with radius R. The segment from the origin to X makes an angle ωt with the x-axis; this angle clearly increases at a constant rate of ω radians per second. The velocity vector of this motion is ωR sin ωt sin ωt v(t)= x′(t)= − = ωR − . ωR cos ωt cos ωt This vector is not constant. however, if you calculate the speed of the point X, you ﬁnd sin ωt v = v(t) = ωR = ωR. cos ωt So while the direction of the velocity vector v(t) is changing all the time, its magnitude is constant. In this parametrization the point X moves along the circle with constant speed v = ωR. 133

54.3. Velocity of the cycloid. Think of the dot X on the wheel in the cycloid example 51.6. We know its position vector and velocity at time t t sin t 1 cos t x (t)= − , x′(t)= − . 1 cos t sin t − The speed with which X traces out the cycloid is

v = x ′(t) = (1 cos t)2 + (sin t)2 − = 1 2 cos t + cos2 t + sin2 t − = 2(1 cos t). − You can use the double angle formula cos 2α = 1 2 sin2 α with α = t to simplify this to − 2 t v = 4 sin2 t = 2 sin . 2 2 The speed of the point X on the cycloid is therefore always between 0 and 2. At times t = 0 and other multiples of 2π we have x′(t)= 0. At these times the point X has come 2 to a stop. At times t = π + 2kπ one has v = 2 and x′(t)=( 0 ), i.e. the point X is moving horizontally to the right with speed 2.

55. Acceleration and Force

Just as the derivative x′(t) of a parametric curve can be interpreted as the velocity vector v(t), the derivative of the velocity vector measures the rate of change with time of the velocity and is called the acceleration of the motion. The usual notation is dv(t) d2x a(t)= v′(t)= = = x ′′(t). dt dt2 Sir Isaac Newton’s law relating force and acceleration via the formula “F = ma” has a vector version. If an object’s motion is given by a parametrized curve x (t) then this motion is the result of a force F being exerted on the object. The force F is given by d2x F = ma = m dt2 where m is the mass of the object. Somehow it is always assumed that the mass m is a positive number.

55.1. How does an object move if no forces act on it? If F (t) = 0 at all times, then, assuming m = 0 it follows from F = ma that a(t) = 0. Since a(t) = v′(t) you conclude that the velocity vector v(t) must be constant, i.e. that there is some ﬁxed vector v such that

x′(t)= v(t)= v for all t. This implies that x(t)= x (0) + tv. So if no force acts on an object, then it will move with constant velocity vector along a straight line (said Newton – Archimedes long before him thought that the object would slow down and come to a complete stop unless there were a force to keep it going.) 134

55.2. Compute the forces acting on a point on a circle. Consider an object moving with constant angular velocity ω on a circle of radius R, i.e. consider x (t) as in example 54.2, R cos ωt cos ωt x (t)= = R . R sin ωt sin ωt Then its velocity and acceleration vectors are sin ωt v(t)= ωR − cos ωt and

2 cos ωt a(t)= v′(t)= ω R − sin ωt − cos ωt = ω2R − sin ωt

cos θ sin θ Since both ( sin θ ) and −cos θ are unit vectors, we see that F the velocity vector changes its direction but not its size: at all θ times you have v = v = ωR. The acceleration also keeps v changing its direction, but its magnitude is always v 2 v2 a = a = ω2R = R = . R R The force which must be acting on the object to make it go through this motion is cos ωt F = ma = mω2R . − sin ωt To conclude this example note that you can write this force as

F = mω2x(t) − which tells you which way the force is directed: towards the center of the circle.

55.3. How does it feel, to be on the Ferris wheel? In other words, which force acts on us if we get carried away by a “liberated ferris wheel,” as in example 51.6?

Well, you get pushed around by a force F , which according to Newton is given by F = ma, where m is your mass and a is your acceleration, which we now compute: C a a(t)= v′(t) X t d 1 cos t = − dt sin t O A sin t = . cos t This is a unit vector: the force that’s pushing you around is con- stantly changing its direction but its strength stays the same. If you remember that t is the angle ∠ACX you see that the force F is always pointed at the center of the wheel: its direction is given by the vector −−→XC. 135

56. Tangents and the unit tangent vector

Here we address the problem of ﬁnding the tangent line at a point on a parametric curve. Let x (t) be a parametric curve, and let’s try to ﬁnd the tangent line at a particular point X0, with position vector x (t0) on this curve. We follow the same strategy as in 1st semester calculus: pick a point Xh on the curve near X0, draw the line through X0 and Xh and let Xh X0. → The line through two points on a curve is often called a secant to the curve. So we are going to construct a tangent to the curve as a limit of secants.

The point X0 has position vector x(t0), the point Xh is at x (t0 + h). Consider the line ℓh parametrized by

x(t0 + h) x(t0) (74) y(s; h)= x(t0)+ s − , h in which s is the parameter we use to parametrize the line.

x′(t0) ℓ ℓh

Xh x(t ) X0 ℓ 0

The line ℓh contains both X0 (set s = 0) and Xh (set s = h), so it is the line through X0 and Xh, i.e. a secant to the curve. Now we let h 0, which gives → def x(t0 + h) x(t0) y(s) = lim y(s; h)= x(t0)+ s lim − = x(t0)+ sx′(t0), h 0 h 0 h → → In other words, the tangent line to the curve x(t) at the point with position vector x(t0) has parametric equation

y(s)= x(t0)+ sx′(t0), and the vector x′(t0) = v(t0) is parallel to the tangent line ℓ. Because of this one calls the vector x′(t0) a tangent vector to the curve. Any multiple λx′(t0) with λ = 0 is still parallel to the tangent line ℓ and is therefore also called a tangent vector.

A tangent vector of length 1 is called a unit tangent vector. If x′(t0) = 0 then there are exactly two unit tangent vectors. They are

v(t0) v(t0) T (t0)= = . ± v(t0) ± v(t0)

56.1. Example. Find Tangent line, and unit tangent vector at x(1), where x (t) is the parametric curve given by t 1 x(t)= , so that x′(t)= . t2 2t 136

Circle with radius 1

1.5 T

1 1

0.5

T −

0.5 1 1

parabola with tangent line parabola with unit tangent vectors

1 Solution: For t = 1 we have x ′(1) = ( 2 ), so the tangent line has parametric equation 1 1 1+ s y(s)= x (1) + sx′(1) = + s = . 1 2 1 + 2s In components one could write this as y1(s)=1+ s, y2(s)=1+2s. After eliminating s you ﬁnd that on the tangent line one has

y2 =1+2s =1+2(y1 1) = 2y1 1. − − 1 The vector x ′(1) = ( 2 ) is a tangent vector to the parabola at x (1). To get a unit tangent vector we normalize this vector to have length one, i.e. we divide i by its length. Thus 1 1 1 √5 T (1) = = 5 √ 2 2 2 2 √5 1 + 2 5 is a unit tangent vector. There is another unit tangent vector, namely 1 √5 T (1) = 5 . − − 2 √5 5 56.2. Tangent line and unit tangent vector to Circle. In example 51.5 and 52.1 we had parametrized the circle and found the velocity vector of this parametrization, cos θ sin θ x(θ)= , x′(θ)= − . sin θ cos θ If we pick a particular value of θ then the tangent line to the circle at x(θ0) has parametric equation cos θ + s sin θ y(s)= x(θ )+ sx′(θ )= 0 0 sin θ s cos θ − This equation completely describes the tangent line, but you can try to write it in a more familiar form as a graph y2 = my1 + n. To do this you have to eliminate the parameter s from the parametric equations

y1 = cos θ + s sin θ, y2 = sin θ s cos θ. − When sin θ = 0 you can solve y1 = cos θ + s sin θ for s, with result y1 cos θ s = − . sin θ 137

So on the tangent line you have

y1 cos θ y2 = sin θ s cos θ = sin θ cos θ − − − sin θ which after a little algebra (add fractions and use sin2 θ + cos2 θ = 1) turns out to be the same as 1 y2 = cot θ y1 + . − sin θ

1 1 sin θ θ π θ 2 −

The tangent line therefore hits the vertical axis when y1 = 0, at height n = 1/ sin θ, and it has slope m = cot θ. − For this example you could have found the tangent line without using any calculus by studying the drawing above carefully.

Finally, let’s ﬁnd a unit tangent vector. A unit tangent is a multiple of x ′(θ) whose sin θ length is one. But the vector x′(θ)= −cos θ already has length one, so the two possible unit vectors are sin θ sin θ T (θ)= x′(θ)= − and T (θ)= . cos θ − cos θ −

57. Sketching a parametric curve

For a given parametric curve, like 1 t2 (75) x(t)= − 3 3t t − you might want to know what the curve looks like. The most straightforward way of getting a picture is to compute x1(t) and x2(t) for as many values of t as you feel like, and then plotting the computed points. This computation is the kind of repetitive task that computers are very good at, and there are many software packages and graphing calculators that will attempt to do the computation and drawing for you. If the vector function has a constant whose value is not (completely) known, e.g. if we wanted to graph the parametric curve 1 t2 (76) x(t)= − 3 (a is a constant) 3at t − then plugging parameter values and plotting the points becomes harder, since the unknown constant a shows up in the computed points. On a graphing calculator you would have to choose different values of a and see what kind of pictures you get (you would expect different pictures for different values of a). 138

In this section we will use the information stored in the derivative x′(t) to create a rough sketch of the graph by hand. Let’s do the speciﬁc curve (75) ﬁrst. The derivative (or velocity vector) is

2t x1′ (t) = 2t x ′(t)= − 2 , so − 2 3 3t x′ (t) = 3(1 t ) − 1 − We see that x′ (t) changes its sign at t = 0, while x′ (t) = 2(1 t)(1 + t) changes its sign 1 2 − twice, at t = 1 and then at t = +1. You can summarize this in a drawing: − +++ + + +− − − − − −− − −

sign of x1′ (t)

− − − − − + ++ + + + + −− − − sign of x2′ (t)

t = 1 t = 0 t = 1 − The arrows indicate the wind direction of the velocity vector x ′(t) for the various values of t. 2 ( − ) 0 2 For instance, when t< 1 you have x1′ (t) > 0 and x2′ (t) < 0, so − ′ x1(t) + 0 that the vector x′(t) = ′ = points in the direction ( 3 ) x2(t) − ? 1 “South-East.” You see that there are three special t values at which x′(t) is either purely horizontal or vertical. Let’s compute 1 x(t) at those values 0 2 t = 1 x( 1) = 2 x′( 1)=( 0 ) − − − − ? 1 1 0 t = 0 x(0) = ( 0 ) x′(0) = ( 3 ) 0 2 2 t = 1 x(1) = ( ) x′(1) = ( − ) ( 0 ) − 2 0 This leads you to the following sketch: -2

If you use a plotting program like Gnuplot you get this picture 139

58. Length of a curve

If you have a parametric curve x(t), a t b, then there is a formula for the length ≤ ≤ of the curve it traces out. We’ll go through a brief derivation of this formula before stating it.

To compute the length of the curve x(t) : { a t b we divide it into lots of short pieces. ≤ ≤ } If the pieces are short enough they will be almost straight line segments, and we know how do com- x(ti 1) pute the length of a line segment. After comput- − ing the lengths of all the short line segments, you x(ti) add them to get an approximation to the length t = a of the curve. As you divide the curve into ﬁner & ﬁner pieces this approximation should get bet- t = b ter & better. You can smell an integral in this description of what’s coming. Here are some more details: Divide the parameter interval into N pieces,

a = t0

x (ti) x(ti 1) . − − Since we are going to take ti 1 ti “very small,” we can use the derivative to approximate − − the distance by

x(ti) x (ti 1) x (ti) x(ti 1)= − − ti ti 1 x′(ti) ti ti 1 , − − ti ti 1 − − ≈ − − − − so that

x(ti) x(ti 1) x ′(ti) ti ti 1 . − − ≈ − − Now add all these distances and you get N b Length polygon x ′(ti) ti ti 1 x′(t) dt. ≈ − − ≈ i=1 t=a This is our formula for the length of a curve. Just in case you think this was a proof, it isn’t! First, we have used the symbol ≈ which stands for “approximately equal,” and we said “very small” in quotation marks, so there are several places where the preceding discussion is vague. But most of all, we can’t prove that this integral is the length of the curve, since we don’t have a deﬁnition of “the length of a curve.” This is an opportunity, since it leaves us free to adopt the formula we found as our formal deﬁnition of the length of a curve. Here goes:

58.1. Definition. If x(t) : a t b is a parametric curve, then its length is { ≤ ≤ } given by b Length = x′(t) dt a provided the derivative x′(t) exists, and provided x ′(t) is a Riemann-integrable function. In this course we will not worry too much about the two caveats about differentiability and integrability at the end of the definition. 140

58.2. Length of a line segment. How long is the line segment AB connecting two points A(a1,a2) and B(b1,b2)? Solution: Parametrize the segment by x(t)= a + t(b a), (0 t 1). − ≤ ≤ Then x′(t) = b a , − and thus 1 1 Length(AB)= x ′(t) dt = b a dt = b a . − − 0 0 In other words, the length of the line segment AB is the distance between the two points A and B. It looks like we already knew this, but no, we didn’t: what this example shows is that the length of the line segment AB as defined in definition 58.1 is the distance between the points A and B. So definition 58.1 gives the right answer in this example. If we had found anything else in this example we would have had to change the definition.

58.3. Perimeter of a circle of radius R. What is the length of the circle of radius R centered at the origin? This is another example where we know the answer in advance. The following computation should give us 2πR or else there’s something wrong with deﬁnition 58.1. We parametrize the circle as follows: x(t)= R cos θi + R sin θj, (0 θ 2π). ≤ ≤ Then

2 x′(θ)= R sin θi + R cos θj, and x ′(θ) = R2 sin θ + R2 cos2 θ = R. − The length of this circle is therefore

2π Length of circle = Rdθ = 2πR. 0 Fortunately we don’t have to fix the definition! And now the bad news: The integral in the definition of the length looks innocent enough and hasn’t caused us any problems in the two examples we have done so far. It is however a reliable source of very difficult integrals. To see why, you must write the integral in terms of the components x1(t),x2(t) of x(t). Since

x1′ (t) 2 2 x′(t)= and thus x ′(t) = x1′ (t) + x2′ (t) x2′ (t) the length of the curve parametrized by x(t) : a t b is { ≤ ≤ } b 2 2 Length = x1′ (t) + x2′ (t) dt. a For most choices of x1(t), x2(t) the sum of squares under the square root cannot be simplified, and, at best, leads to a difficult integral, but more often to an impossible integral. But, chin up, sometimes, as if by a miracle, the two squares add up to an expression whose square root can be simplified, and the integral is actually not too bad. Here is an example: 141

58.4. Length of the Cycloid. After getting in at the bottom of a liberated ferris wheel we are propelled through the air along the cycloid whose parametrization is given in example 51.6, θ sin θ x(θ)= − . 1 cos θ − How long is one arc of the Cycloid?

Solution: Compute x ′(θ) and you ﬁnd 1 cos θ x′(θ)= − sin θ so that x′(θ) = (1 cos θ)2 + (sin θ)2 = √2 2 cos θ. − − This doesn’t look promising (this is the function we must integrate!), but just as in example 54.3 we can put the double angle formula cos θ = 1 2 sin2 θ to our advantage: − 2

2 θ θ x ′(θ) = √2 2 cos θ = 4 sin = 2 sin . − 2 2 We are concerned with only one arc of the Cycloid, so we have 0 θ< 2π, which implies ≤ 0 θ π, which in turn tells us that sin θ > 0 for all θ we are considering. Therefore ≤ 2 ≤ 2 the length of one arc of the Cycloid is 2π Length = x ′(θ) dθ 0 2π θ = 2 sin dθ 2 0 2π θ = 2 sin dθ 2 0 θ 2π = 4 cos − 2 0 = 8. To visualize this answer: the height of the cycloid is 2 (twice the radius of the circle), so the length of one arc of the Cycloid is four times its height (Look at the drawing on page 126.)

59. The arclength function

If you have a parametric curve x (t) and you pick a particular point on this curve, say, the point corresponding to X(t) parameter value t0, then one deﬁnes the arclength function X(t0) s(t) (starting at t0) to be t (77) s(t)= x′(τ) dτ t0 Thus s(t) is the length of the curve segment x(τ) : t0 τ t . (τ is a dummy variable.) { ≤ ≤ } If you interpret the parametric curve x (t) as a description of the motion of some object, then the length s(t) of the curve x(τ) : t0 τ t is the distance traveled by { ≤ ≤ } the object since time t0. If you diﬀerentiate the distance traveled with respect to time you should get the speed, and indeed, by the Fundamental Theorem of Calculus one has d t s′(t)= x′(τ) dτ = x ′(t) , dt t0 142 which we had called the speed v(t) in 54. § 60. Graphs in Cartesian and in Polar Coordinates

Cartesian graphs. Most of ﬁrst-semester-calculus deals with a particular kind of curve, namely, the graph of a function, “y = f(x)”. You can regard such a curve as a special kind of parametric curve, where the parametrization is t x(t)= f(t) and we switch notation from “(x,y)” to “(x1,x2).” For this special case the velocity vector is always given by 1 x ′(t)= , f ′(t) the speed is v(t)= x ′(t) = 1+ f (t)2, ′ and the length of the segment between t = a and t = b is b 2 Length = 1+ f ′(t) dt. a Polar graphs. Instead of choosing Cartesian coordinates P (x ,x ) one can consider so-called Polar Coordinates in the plane. r 1 2 θ We have seen these before in the section on complex numbers: to specify the location of a point in the plane you can give its x1,x2 coordinates, but you could also give the absolute value and argument of the complex number x1 + ix2 (see 24.) Or, to say it without mentioning complex numbers, you can § say where a point P in the plane is by saying (1) how far it is from the origin, and (2) how large the angle between the line segment OP and a ﬁxed half line (usually the positive x-axis) is. The Cartesian coordinates of a point with polar coordinates (r, θ) are

(78) x1 = r cos θ, x2 = r sin θ, or, in our older notation, x = r cos θ, y = r sin θ. These are the same formulas as in 24, where we had “ r = z and θ = arg z.” § | | Often a curve is given as a graph in polar coordi- x2 nates, i.e. for each angle θ there is one point (X) on the curve, and its distance r to the origin is some function X f(θ) of the angle. In other words, the curve consists of all points whose polar coordinates satisfy the equation r = f(θ) r = f(θ). You can parametrize such a curve by r cos θ f(θ) cos θ θ (79) x(θ)= = . x r sin θ f(θ) sin θ 1 or, x(θ)= f(θ) cos θi + f(θ) sin θj. You can apply the formulas for velocity, speed and arclength to this parametrization, but instead of doing the straightforward calculation, let’s introduce some more notation. For any angle θ we deﬁne the vector cos θ u (θ)= = cos θi + sin θj. sin θ 143

The derivative of u is sin θ u ′(θ) u ′(θ)= − = sin θi + cos θj. cos θ − u (θ) The vectors u (θ) and u ′(θ) are perpendicular unit vectors. θ Then we have x(θ)= f(θ)u (θ), so by the product rule one has

x′(θ)= f ′(θ)u (θ)+ f(θ)u ′(θ).

x′(θ)

f(θ)u ′(θ) u ′(θ) ψ x′(θ) u (θ) ψ

x′(θ) f ′(θ)u (θ)

Since u (θ) and u ′(θ) are perpendicular unit vectors this implies

v(θ)= x ′(θ) = f (θ)2 + f(θ)2. ′ The length of the piece of the curve between polar angles α and β is therefore β 2 2 (80) Length = f ′(θ) + f(θ) dθ. α You can also read oﬀ that the angle ψ between the radius OX and the tangent to the curve satisﬁes f(θ) tan ψ = . f ′(θ)

61. PROBLEMS

SKETCHING PARAMETRIZED CURVES

3t + 2 Sketch the curves which are traced 464. x(t)= 3t + 2 out by the following parametrizations. Describe the motion (is the curve you t e 465. x(t)= draw traced out once or several times? et In which direction?) et In all cases the parameter is allowed 466. x(t)= to take all values from to . t −∞ ∞ If a curve happens to be the graph of et 467. x(t)= t some function x = f(x ) (or y = f(x) e− 2 1 if you prefer), then ﬁnd the function t f( ). 468. x(t)= t2 Is there a geometric interpretation sin t of the parameter as an angle, or a dis- 469. x(t)= tance, etc? t 1 t sin t 463. x(t)= − 470. x(t)= 2 t cos 2t − 144

sin 25t 471. x(t)= 476. Group problem. A circle of radius cos 25t r > 0 rolls on the outside of the unit 1 + cos t circle. X is a point on the rolling circle 472. x(t)= 1 + sin t (These curves are called epicycloids.) 2 cos t 473. x(t)= 477. Group problem. A circle of radius sin t 0 0 from the cen- 479. X is the point where the tangent line ter of the circle (the case a = 1 gives the at x (θ) to the helix of example 51.7 in- cycloid.) tersects the x1x2 plane.

PRODUCT RULES

480. Group problem. If a moving object has position vector x(t) at time t, and if it’s speed is constant, then show that the acceleration vector is always perpendicular to the velocity vector. [Hint: differentiate v2 = v v with respect to time and use some of the product rules from 53.] § 481. Group problem. If a charged particle moves in a magnetic field B , then the laws of electromagnetism say that the magnetic field exerts a force on the particle and that this force is given by the following miraculous formula: F = qv B . × where q is the charge of the particle, and v is its velocity. Not only does the particle know calculus (since Newton found F = ma), it also knows vector geometry! Show that even though the magnetic field is pushing the particle around, and even though its velocity vector may be changing with time, its speed v = v remains constant. 482. Group problem. Newton’s law of gravitation states that the Earth pulls any object of mass m towards its center with a force inversely proportional to the squared distance of the object to the Earth’s center. (i) Show that if the Earth’s center is the origin, and r is the position vector of the object of mass m, then the gravitational force is given by r F = C (C is a positive constant.) − r 3 [No calculus required. You are supposed to check that this vector satisfies the description in the beginning of the problem, i.e. that it has the right length and direction.] (ii) If the object is moving, then its angular momentum is defined in physics books by the formula L = mr v. Show that, if the Earth’s gravitational field is the only force × 145

acting on the object, then its angular momentum remains constant. [Hint: you should diﬀerentiate L with respect to time, and use a product rule.]

CURVE SKETCHING, USING THE TANGENT VECTOR

483. Consider a triangle ABC and let a,b as a Bezier curve) and c be the position vectors of A, B and x (t)=(1 t)3a+3t(1 t)2b+3t2(1 t)c+t3d C. − − − where 0 t 1. (i) Show that the parametric curve given ≤ ≤ by Compute x(0), x(1), x ′(0), and x ′(1). x(t)=(1 t)2a + 2t(1 t)b + t2c, − − The characters in most fonts (like goes through the points A and C, and the fonts used for these notes) are made that at these points it is tangent to the up of lots of Bezier curves. sides of the triangle. Make a drawing. (ii) At which point on this curve is the tangent parallel to the side AC of the triangle?

484. Let a,b,c, d be four given vectors. Consider the parametric curve (known

485. Sketch the following curves by ﬁnding all points at which the tangent is either horizontal or vertical (in these problems, a is a positive constant.)

2 i 1 t ii sin t ( ) x(t)= − 2 ( ) x(t)= t + 2t sin 2t 2 iii cos t iv 1 t ( ) x (t)= ( ) x(t)= − 3 sin 2t 3at t − 2 v 1 t vi cos 2t ( ) x(t)= − 3 ( ) x (t)= 3at + t sin 3t t/(1 + t2) t2 (vii) x(t)= (viii) x(t)= t2 sin t 1+ t2 (ix) x (t)= 2t4

LENGTHS OF CURVES

486. Find the length of each of the following curve segments. An “ ” indicates a diﬃcult but possible integral which you should do; “ ” indicates that the resulting integral cannot reasonably be done with the methods explained in this course – you may leave an integral 146

in your answer after simplifying it as much as you can. All other problems lead to integrals that shouldn’t be too hard. R(θ sin θ) i − ( ) The cycloid x (θ)= R(1 cos θ) , with 0 θ 2π. − ≤ ≤ cos t (ii) The ellipse x(t)= with 0 t 2π. A sin t ≤ ≤ t (iii) The parabola x (t)= with 0 t 1. t2 ≤ ≤ t (iv) The Sine graph x (t)= with 0 t π. sin t ≤ ≤ cos t + t sin t (v) The evolute of the circle x = (with 0 t L). sin t t cos t ≤ ≤ − x x e + e− (vi) The Catenary, i.e. the graph of y = cosh x = for a x a. 2 − (vii) The Cardioid, which in polar coordinates is given by r = 1 + cos θ, ( θ < π), so | | (1 + cos θ) cos θ x(θ)= . (1 + cos θ) sin θ cos θ (viii) The Helix from example 51.7, x(θ)= sin θ , 0 θ 2π.   ≤ ≤ aθ   487. Below are a number of parametrized curves. For each of these curves find all points with horizontal or vertical tangents; also find all points for which the tangent is parallel to the diagonal. Finally, find the length of the piece of these curves corresponding to the indicated parameter interval (I tried hard to find examples where the integral can be done). t1/3 9 t5/3 (i) x(t)= − 20 0 t 1 t ≤ ≤ t2 (ii) x(t)= 1 t 2 t2√t ≤ ≤ t2 (iii) x(t)= 0 t √3 t t3/3 ≤ ≤ − 8 sin t π (iv) x(t)= t 7t sin t cos t | |≤ 2 − t (v) Groupproblemx (t)= 0 t 1 √1+ t ≤ ≤ (The last problem is harder, but it can be done. In all the other ones the quantity under the square root that appears when you set up the integral for the length of the curve is a perfect square.)

488. Consider the polar graph r = ekθ, with <θ< , where k is a positive constant. −∞ ∞ This curve is called the logarithmic spiral. (i) Find a parametrization for the polar graph of r = ekθ.

(ii) Compute the arclength function s(θ) starting at θ0 = 0. (iii) Show that the angle between the radius and the tangent is the same at all points on the logarithmic spiral. (iv) Which points on this curve have horizontal tangents? 147

489. Group problem. The Archimedean spiral is the polar graph of r = θ, where θ 0. ≥ (i) Which points on the part of the spiral with 0 <θ<π have a horizontal tangent? Which have a vertical tangent? (ii) Find all points on the whole spiral (allowing all θ> 0) which have a horizontal tangent. (iii) Show that the part of the spiral with 0 <θ<π is exactly as long as the piece of the 1 2 parabola y = 2 x between x = 0 and x = π. (It is not impossible to compute the lengths of both curves, but you don’t have to to answer this problem!)

KEPLER’s LAW’s

Kepler’s ﬁrst law: Planets move in a plane in an ellipse with the sun at one focus.

Kepler’s second law: The position vector from the sun to a planet sweeps out area at a constant rate.

Kepler’s third law: The square of the period of a planet is proportional to the cube of its mean distance from the sun. The mean distance is the average of the closest distance and the furthest distance. The period is the time required to go once around the sun.

Let p = xi + yj + zk be the position of a planet in space where x, y and z are all function of time t. Assume the sun is at the origin. Newton’s law of gravity implies that d2p p = α (1) dt2 p 3 || || where α is GM, G is a universal gravitational constant and M is the mass of the sun. − It does not depend on the mass of the planet. First let us show that planets move in a plane. By the product rule d dp dp dp d2p (p )=( )+(p ) (2) dt × dt dt × dt × dt2 By (1) and the fact that the cross product of parallel vectors is 0 the right hand side of (2) is 0. It follows that there is a constant vector c such that at all times dp p = c (3) × dt Thus we can conclude that both the position and velocity vector lie in the plane with normal vector c. Without loss of generality we assume that c = βk for some scaler β and p = xi + yj. Let x = r cos(θ) and y = r sin(θ) where we consider r and θ as functions of t. If we calculate the derivative of p we get dp dr dθ dr dθ = cos(θ) r sin(θ) i + sin(θ)+ r cos(θ) j (4) dt dt − dt dt dt Since p dp = βk we have × dt dr dθ dr dθ r cos(θ) sin(θ)+ r cos(θ) r sin(θ) cos(θ) r sin(θ) = β (5) dt dt − dt − dt After multiplying out and simplifying this reduces to dθ r2 = β (6) dt The area swept out from time t0 to time t1 by a curve in polar coordinates is 1 t1 dθ A = r2 dt (7) 2 dt t0 By (6) A is proportional to t1 t0. This is Kepler’s second law. − 148

We will now prove Kepler’s third law for the special case of a circle. So let T be the time it takes the planet to go around the sun one time and let r be its distance from the sun. We will show that T 2 (2π)2 = (8) r3 − α The second law implies that θ(t) is a linear function of t and so in fact dθ 2π = (9) dt T dr Since r is constant we have that dt = 0 and so (4) simplifies to dp 2π 2π = r sin(θ) i + r cos(θ) j (10) dt − T T Differentiating once more we get 2 d p 2π 2 2π 2 2π 2 = r cos(θ) i + r sin(θ) j = p (11) dt2 − T − T − T Noting that r = p and using (1) we get || || α 2π 2 = (12) r3 − T from which (8) immediately follows. Complete derivations of the three laws from Newton’s law of gravity can be found in T.M.Apostal, Calculus vol I , Blaisdel(1967), p.545-548. Newton deduced the law of gravity from Kepler’s laws. The argument can be found in L.Bers, Calculus vol II , Holt,Rinhart,and Winston(1969), p.748-754. The planet earth is 93 million miles from the sun. The year has 365 days. The moon is 250,000 miles from the earth and circles the earth once every 28 days. The earth’s diameter is 7850 miles. In the first four problems you may assume orbits are circular. Use only the data in this paragraph.

490. The former planet Pluto takes 248 years to orbit the sun. How far is Pluto from the sun? Mercury is 36 million miles from the sun. How many (Earth) days does it take for Mercury to complete one revolution of the sun? 491. Russia launched the first orbital satelite in 1957. Sputnik orbited the earth every 96 minutes. How high off the surface of the earth was this satelite? 492. A communication satellite is to orbit the earth around the equator at such a distance so as to remain above the same spot on the earth’s surface at all times. What is the distance from the center of the earth such a satellite should orbit? 493. Find the ratio of the masses of the sun and the earth. 494. The Kmart7 satellite is to be launched into polar earth orbit by firing it from a large cannon. This is possible since the satellite is very small, consisting of a single blinking blue light. Polar orbit means that the orbit passes over both the north and south poles. Let p(t) be the point on the earth’s surface at which the blinking blue light is directly overhead at time t. Find the largest orbit that the Kmart7 can have so that every person on earth will be within 1000 miles of p(t) at least once a day. You may assume that the satellite orbits the earth exactly n times per day for some integer n.

2A 495. Let A be the total area swept out by an elliptical orbit. Show that β = T . 496. Let E be an ellipse with one of the focal points f. Let d be the minimum distance from some point of the ellipse to f and let D be the maximum distance. In terms of d and D only what is the area of the ellipse E? 149

Hint: The area of an ellipse is πab where a is its minimum radius and b its maximum radius (both from the center of the ellipse). If f1 and f2 are the focal points of E then the sum of the distances from f1 to p and f2 to p is constant for all points p on E. 497. Halley’s comet orbits the sun every 77 years. Its closest approach is 53 million miles. What is its furthest distance from the sun? What is the maximum speed of the comet and what is the minimum speed? 150

GNU Free Documentation License

Version 1.3, 3 November 2008 Copyright c 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc.

http://fsf.org/

Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.

Preamble

The purpose of this License is to make a manual, textbook, or other functional and useful document “free” in the sense of freedom: to assure everyone the eﬀective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modiﬁcations made by others. This License is a kind of “copyleft”, which means that derivative works of the document must themselves be free in the same sense. It complements the GNU General Public License, which is a copyleft license designed for free software. We have designed this License in order to use it for manuals for free software, because free software needs free documentation: a free program should come with manuals providing the same freedoms that the software does. But this License is not limited to software manuals; it can be used for any textual work, regardless of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is instruction or reference.

1. APPLICABILITY AND DEFINITIONS

This License applies to any manual or other work, in any medium, that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that work under the conditions stated herein. The “Document”, below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as “you”. You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law. A “Modified Version” of the Document means any work containing the Document or a portion of it, either copied verbatim, or with modifications and/or translated into another language. A “Secondary Section” is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document’s overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (Thus, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them. The “Invariant Sections” are certain Secondary Sections whose titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none. The “Cover Texts” are certain short passages of text that are listed, as Front-Cover Texts or Back-Cover Texts, in the notice that says that the Document is released under this License. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most 25 words. A “Transparent” copy of the Document means a machine-readable copy, represented in a format whose specifi- cation is available to the general public, that is suitable for revising the document straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not “Transparent” is called “Opaque”. Examples of suitable formats for Transparent copies include plain ASCII without markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only. The “Title Page” means, for a printed book, the title page itself, plus such following pages as are needed to hold, legibly, the material this License requires to appear in the title page. For works in formats which do not have any title page as such, “Title Page” means the text near the most prominent appearance of the work’s title, preceding the beginning of the body of the text. 151

The “publisher” means any person or entity that distributes copies of the Document to the public. A section “Entitled XYZ” means a named subunit of the Document whose title either is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in another language. (Here XYZ stands for a specific section name mentioned below, such as “Acknowledgements”, “Dedications”, “Endorsements”, or “History”.) To “Preserve the Title” of such a section when you modify the Document means that it remains a section “Entitled XYZ” according to this definition. The Document may include Warranty Disclaimers next to the notice which states that this License applies to the Document. These Warranty Disclaimers are considered to be included by reference in this License, but only as regards disclaiming warranties: any other implication that these Warranty Disclaimers may have is void and has no effect on the meaning of this License.

2. VERBATIM COPYING

You may copy and distribute the Document in any medium, either commercially or noncommercially, provided that this License, the copyright notices, and the license notice saying this License applies to the Document are reproduced in all copies, and that you add no other conditions whatsoever to those of this License. You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute. However, you may accept compensation in exchange for copies. If you distribute a large enough number of copies you must also follow the conditions in section 3. You may also lend copies, under the same conditions stated above, and you may publicly display copies.

3. COPYING IN QUANTITY

If you publish printed copies (or copies in media that commonly have printed covers) of the Document, numbering more than 100, and the Document’s license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both covers must also clearly and legibly identify you as the publisher of these copies. The front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects. If the required texts for either cover are too voluminous to fit legibly, you should put the first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto adjacent pages. If you publish or distribute Opaque copies of the Document numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public- standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public. It is requested, but not required, that you contact the authors of the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document.

4. MODIFICATIONS

You may copy and distribute a Modified Version of the Document under the conditions of sections 2 and 3 above, provided that you release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version:

A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission. B. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modifications in the Modified Version, together with at least five of the principal authors of the Document (all of its principal authors, if it has fewer than five), unless they release you from this requirement. C. State on the Title page the name of the publisher of the Modified Version, as the publisher. D. Preserve all the copyright notices of the Document. E. Add an appropriate copyright notice for your modifications adjacent to the other copyright notices. F. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modified Version under the terms of this License, in the form shown in the Addendum below. G. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Document’s license notice. H. Include an unaltered copy of this License. I. Preserve the section Entitled “History”, Preserve its Title, and add to it an item stating at least the title, year, new authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled “History” in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence. J. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be placed in the “History” section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission. K. For any section Entitled “Acknowledgements” or “Dedications”, Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor acknowledgements and/or dedications given therein. 152

L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles. M. Delete any section Entitled “Endorsements”. Such a section may not be included in the Modiﬁed Version. N. Do not retitle any existing section to be Entitled “Endorsements” or to conﬂict in title with any Invariant Section. O. Preserve any Warranty Disclaimers.

If the Modiﬁed Version includes new front-matter sections or appendices that qualify as Secondary Sections and contain no material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modiﬁed Version’s license notice. These titles must be distinct from any other section titles.

You may add a section Entitled “Endorsements”, provided it contains nothing but endorsements of your Modiﬁed Version by various parties—for example, statements of peer review or that the text has been approved by an organization as the authoritative deﬁnition of a standard.

You may add a passage of up to ﬁve words as a Front-Cover Text, and a passage of up to 25 words as a Back- Cover Text, to the end of the list of Cover Texts in the Modiﬁed Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one.

The author(s) and publisher(s) of the Document do not by this License give permission to use their names for publicity for or to assert or imply endorsement of any Modiﬁed Version.

5. COMBINING DOCUMENTS

You may combine the Document with other documents released under this License, under the terms defined in section 4 above for modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, unmodified, and list them all as Invariant Sections of your combined work in its license notice, and that you preserve all their Warranty Disclaimers.

The combined work need only contain one copy of this License, and multiple identical Invariant Sections may be replaced with a single copy. If there are multiple Invariant Sections with the same name but diﬀerent contents, make the title of each such section unique by adding at the end of it, in parentheses, the name of the original author or publisher of that section if known, or else a unique number. Make the same adjustment to the section titles in the list of Invariant Sections in the license notice of the combined work.

In the combination, you must combine any sections Entitled “History” in the various original documents, forming one section Entitled “History”; likewise combine any sections Entitled “Acknowledgements”, and any sections Entitled “Dedications”. You must delete all sections Entitled “Endorsements”.

6. COLLECTIONS OF DOCUMENTS

You may make a collection consisting of the Document and other documents released under this License, and replace the individual copies of this License in the various documents with a single copy that is included in the collection, provided that you follow the rules of this License for verbatim copying of each of the documents in all other respects.

You may extract a single document from such a collection, and distribute it individually under this License, provided you insert a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document.

7. AGGREGATION WITH INDEPENDENT WORKS

A compilation of the Document or its derivatives with other separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an “aggregate” if the copyright resulting from the compilation is not used to limit the legal rights of the compilation’s users beyond what the individual works permit. When the Document is included in an aggregate, this License does not apply to the other works in the aggregate which are not themselves derivative works of the Document.

If the Cover Text requirement of section 3 is applicable to these copies of the Document, then if the Document is less than one half of the entire aggregate, the Document’s Cover Texts may be placed on covers that bracket the Document within the aggregate, or the electronic equivalent of covers if the Document is in electronic form. Otherwise they must appear on printed covers that bracket the whole aggregate.

8. TRANSLATION

Translation is considered a kind of modiﬁcation, so you may distribute translations of the Document under the terms of section 4. Replacing Invariant Sections with translations requires special permission from their copyright holders, but you may include translations of some or all Invariant Sections in addition to the original versions of these Invariant Sections. You may include a translation of this License, and all the license notices in the Document, and any Warranty Disclaimers, provided that you also include the original English version of this License and the original versions of those notices and disclaimers. In case of a disagreement between the translation and the original version of this License or a notice or disclaimer, the original version will prevail.

If a section in the Document is Entitled “Acknowledgements”, “Dedications”, or “History”, the requirement (section 4) to Preserve its Title (section 1) will typically require changing the actual title.

9. TERMINATION 153

You may not copy, modify, sublicense, or distribute the Document except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense, or distribute it is void, and will automatically terminate your rights under this License. However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation. Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice. Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, receipt of a copy of some or all of the same material does not give you any rights to use it.

10. FUTURE REVISIONS OF THIS LICENSE

The Free Software Foundation may publish new, revised versions of the GNU Free Documentation License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. See http://www.gnu.org/copyleft/. Each version of the License is given a distinguishing version number. If the Document specifies that a particular numbered version of this License “or any later version” applies to it, you have the option of following the terms and conditions either of that specified version or of any later version that has been published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation. If the Document specifies that a proxy can decide which future versions of this License can be used, that proxy’s public statement of acceptance of a version permanently authorizes you to choose that version for the Document.

11. RELICENSING

“Massive Multiauthor Collaboration Site” (or “MMC Site”) means any World Wide Web server that publishes copyrightable works and also provides prominent facilities for anybody to edit those works. A public wiki that anybody can edit is an example of such a server. A “Massive Multiauthor Collaboration” (or “MMC”) contained in the site means any set of copyrightable works thus published on the MMC site. “CC-BY-SA” means the Creative Commons Attribution-Share Alike 3.0 license published by Creative Commons Corporation, a not-for-proﬁt corporation with a principal place of business in San Francisco, California, as well as future copyleft versions of that license published by that same organization. “Incorporate” means to publish or republish a Document, in whole or in part, as part of another Document. An MMC is “eligible for relicensing” if it is licensed under this License, and if all works that were ﬁrst published under this License somewhere other than this MMC, and subsequently incorporated in whole or in part into the MMC, (1) had no cover texts or invariant sections, and (2) were thus incorporated prior to November 1, 2008. The operator of an MMC Site may republish an MMC contained in the site under CC-BY-SA on the same site at any time before August 1, 2009, provided the MMC is eligible for relicensing.